Electrocardiogram‐Based Predictors of Sudden Cardiac Death in Patients With Coronary Artery Disease

Abstract

Current recommendations on which patients with coronary artery disease (CAD) should be offered an implantable cardioverter defibrillator for the primary prevention of sudden cardiac death (SCD) rely heavily on the presence of depressed left ventricular ejection fraction. Because the majority of SCD victims with CAD have preserved left ventricular function, additional cardiac investigations are likely to play an increasing role in the risk stratification of CAD patients. A number of studies have demonstrated that certain parameters on the traditional 12‐lead electrocardiogram (ECG) and other ECG‐based investigations (such as signal‐averaged ECG, heart rate turbulence, T‐wave alternans) can provide important information on the underlying cardiac substrate abnormality that may predispose to ventricular arrhythmias and SCD. This article reviews some of the evidence for these ECG‐based tests as predictors of SCD in patients with CAD and addresses their advantages and limitations. © 2011 Wiley Periodicals, Inc.

The author has no funding, financial relationships, or conflicts of interest to disclose.

Introduction

The majority of sudden cardiac death (SCD) victims have underlying coronary artery disease (CAD), which may or may not have been previously known to the victim.1 Significantly impaired left ventricular ejection fraction (LVEF) is an established predictor of SCD and forms the basis of current guidelines on which patients should be offered an implantable cardioverter defibrillator (ICD) for primary prevention. However, less than one third of all SCD victims have significantly impaired LVEF.1 Consequently, a large number of patients with CAD remain at risk of SCD, and improvements in current methods of risk stratification are likely to have a significant impact on the incidence of SCD. At present, there is considerable interest among clinicians, researchers, and healthcare providers on improved tests for risk stratification of SCD and more precise methods of determining which patients would benefit most from prophylactic ICD insertion. Two groups of patients could benefit from such improved methods. The first group is patients with impaired LVEF who fulfill current guidelines for primary prevention ICDs; long‐term follow‐up data of ICD recipients has shown that a significant proportion of patients never require any ICD therapy. Conversely, as many as 25% experience an inappropriate ICD shock, which itself has associated morbidity and mortality.2, 3 Thus, more precise methods of risk stratification of SCD could improve our current selection criteria for selecting prophylactic ICD recipients. The second group is patients with relatively preserved LVEF (>35%) who are not included in current guidelines. Improved, noninvasive methods of risk stratification may allow this group of patients (who make up the majority of SCD victims) to be investigated and incorporated into future guidelines.

A number of studies have demonstrated that certain parameters on the 12‐lead ECG and other ECG‐based investigations can provide important information on the underlying cardiac substrate abnormality that may predispose to ventricular arrhythmias and SCD. This article reviews some of these ECG‐based predictors of SCD in patients with CAD, which may have a role to play in future algorithms designed to enhance risk stratification and determine who should be offered a prophylactic ICD.

Clues From the Standard 12‐Lead ECG

Several parameters on the standard 12‐lead ECG have been shown to provide prognostic information in patients with CAD. These include the presence of bundle branch block, prolonged QRS duration, left ventricular hypertrophy, and QT dispersion. Many of these parameters are also markers of significant left ventricular (LV) dysfunction rather than specific predictors of SCD, and therefore their sole use in predicting which patients are at highest risk of dying suddenly is limited. However, it is useful for clinicians to be aware of these parameters and their clinical significance, as their presence may sway one to perform additional cardiac investigations to refine the risk of SCD.

QRS Duration

Increased QRS duration has been known for some time to be associated with increased mortality.4, 5 However, there is little evidence that increased QRS duration is associated with an increased risk of sudden cardiac death in patients with CAD. The Multi‐Centre Unsustained Tachycardial Trial (MUSTT) investigators found that although left bundle branch block and nonspecific intraventricular conduction delay were associated with increased total mortality, there was no significant link between these parameters and inducible monomorphic ventricular tachycardia (VT).6 In the PainFree RX II trial, QRS duration did not predict the delivery of appropriate therapies for VT or ventricular fibrillation (VF) in 431 patients with CAD who received an ICD for either primary or secondary prevention.7 Increased QRS duration has, however, been demonstrated to predict SCD in hypertensive patients undergoing intensive medical therapy, many of whom may also have underlying CAD.8

QT Interval and QT Dispersion

Early studies investigating the possible association between a prolonged QT interval (corrected for heart rate) and mortality after myocardial infarction reported conflicting results.9, 10 This may be due to several reasons, including wide overlap in QT interval measurements between subjects with and without events, and difficulties in measuring the QT interval accurately in some leads due to T‐U wave abnormalities. Similarly, other investigators who have examined the relationship between QT dispersion, defined as the difference between the longest and shortest QT interval on the standard 12‐lead ECG (excluding aVR), and SCD have reported varying results.11, 12, 13 The potential association between QT indices and SCD may be more complicated than initially realized as genetic and racial factors may also be relevant. For example, recent studies analyzing single nucleotide polymorphisms of the nitric oxide synthase adaptor protein 1 (NOS1AP), which is associated with the QT interval in white adults, have demonstrated a link with SCD.14, 15 In the Oregon Sudden Unexpected Death Study, 373 patients with CAD who had prior 12‐lead ECGs and died suddenly and unexpectedly were compared with 309 controls with CAD who had not died suddenly.16 The investigators found that the mean corrected QT interval was significantly longer in the cases than controls and demonstrated that QT‐prolonging drugs and idiopathic abnormal QT prolongation were significantly associated with SCD. It is therefore likely that other determinants of QT prolongation, including as yet identified genomic factors, increase the risk of SCD in patients with CAD.

Fragmented QRS Complexes and QRS Scores

The presence of a fragmented QRS complex (fQRS) on the 12‐lead ECG has been described as a marker of abnormal ventricular depolarization and demonstrated to be a predictor of mortality and SCD.17 Fragmented QRS complexes include various RSR′ patterns, with or without QRS duration <120 ms and probably represent conduction delay caused by myocardial scar in patients with CAD. fQRS is a simple, inexpensive, and easily measured ECG parameter that may be of value in determining the risk of SCD and guiding prophylactic ICD insertion in high‐risk individuals. An example of fragmented QRS complexes on the 12‐lead ECG is shown in Figure 1.

Figure 1.

Example of fragmented QRS complexes from the 12‐lead electrocardiogram of a patient with previous myocardial infarction and depressed left ventricular function. Fragmentation is shown by the presence of more than 2 R′ in 2 contiguous inferior leads (III and aVF), even in the presence of typical left bundle branch block (QRS duration, 120 ms), indicating the presence of inferior scar.

Recently, Strauss et al described a new QRS scoring system that correlates with the presence of scar in ischemic and nonischemic cardiomyopathy patients.18 The QRS scoring system involves measurement of Q‐, R‐ and S‐wave amplitudes, durations, amplitude ratios, and notches in 10 of the 12 standard ECG leads (excluding leads III and aVR). These investigators subsequently applied QRS scoring to 797 patients from the ICD arm of the Sudden Cardiac Death in Heart Failure Trial, and found that a higher QRS score predicted increased rates of ventricular arrhythmias during follow‐up.19 Thus, like the presence of fragmented QRS complexes, the QRS score appears to provide additional prognostic information that may have a role in risk‐stratifying algorithms and selection of ICD recipients.

Early Repolarization

The presence of early repolarization of the QRS complex, previously regarded as a benign and a normal variant, is now recognized to have prognostic significance in some individuals and may have a potential role to play in the risk prediction of ventricular arrhythmias. Early repolarization, defined as elevation of the QRS‐ST junction (J point) by 0.1 mV in at least 2 leads (other than V1 to V3), has been demonstrated to occur in patients with idiopathic VF.20, 21 Whether early repolarization is a useful parameter to predict SCD in patients with CAD remains to be determined, although its simplicity and ease of measurement make it an attractive parameter for further investigation.

Signal‐Averaged ECG

The signal‐averaged electrocardiogram (SAECG) is a test to identify the presence of ventricular late potentials (VLPs), which represent slowed conduction through a diseased myocardium due to the presence of fibrosis or scar that may form the substrate for ventricular arrhythmias. VLPs cannot usually be detected on conventional 12‐lead ECG in view of their very low amplitude (microvolt range), and so an amplified, high‐resolution ECG recording is required for their identification. The presence of VLPs in patients with CAD has been demonstrated to predict inducibility of sustained VT during electrophysiological testing, with a sensitivity of 87% and specificity of 65%.22 Combining the SAECG with ejection fraction improved the predictive accuracy. Similarly, a substudy of patients in the MUSTT trial with CAD found that the combination of SAECG with LVEF <40% was predictive of arrhythmic or cardiac death at both 2‐ and 5‐year follow‐up.23 The greatest predictor of an arrhythmic event was filtered QRS duration of >114 ms in patients with an LVEF <30%.

A number of studies have demonstrated that the SAECG is predictive of arrhythmic events and SCD in patients following myocardial infarction.24, 25 However, although the negative predictive value of SAECG is high (>95%), the positive predictive accuracy is much lower, thus decreasing its usefulness as a single variable to identify high‐risk patients. With the increasing use of primary percutaneous coronary intervention (PCI), the prognostic value of the SAECG has become less clear. Bauer et al performed SAECGs in 968 patients following acute myocardial infarction, 91% of whom underwent PCI, and found that the presence of VLPs was not significantly associated with cardiac death or a serious arrhythmic event during a median follow‐up of 34 months.26

Holter Analysis

Detection of Ventricular Arrhythmias

As with SAECG, much of the data on the usefulness of Holter analysis in risk prediction of SCD is obtained from data after myocardial infarction. Early studies reported that the detection of ventricular arrhythmias, most often nonsustained VT or frequent premature ventricular complexes (PVCs), on Holter monitoring is predictive of serious arrhythmic events and death.27, 28 A more recent study of 2130 patients following myocardial infarction, in which 70% underwent coronary revascularization, demonstrated that nonsustained VT remained an independent predictor of SCD after adjustment for age, diabetes, and LVEF and was especially useful in patients with LVEF >35%.29 In a less acute setting of 867 patients referred for ambulatory 24‐hour ECG monitoring, high resting heart rate and the presence of PVCs on 24‐hour Holter monitoring appear to be independently associated with ventricular arrhythmias.30 Furthermore, a recent case‐control analysis of the 49 (out of 1649) patients who died suddenly in the Cardiovascular Health Study suggested that Holter monitoring may provide useful information on SCD risk in older patients (>65 years) with CAD risk factors and stroke.31

Although some useful data can be obtained from conventional 24‐hour Holter monitoring, its routine use for assessing the risk of SCD in patients with CAD is not recommended in view of its low sensitivity and specificity. Recent advances in Holter‐based technology have allowed for additional parameters to be obtained from ambulatory ECG‐recordings that may provide more prognostic information. These parameters, such as heart rate variability and heart rate turbulence, represent changes in cardiac autonomic tone that occur following myocardial infarction.

Heart Rate Variability

Heart rate variability (HRV) can be assessed using various methods by measuring ECG recordings over a short (2–30 minutes) or longer (24 hour) periods. Decreased HRV appears to be associated with increased ventricular arrhythmias and mortality.32, 33 In the Multicenter Postinfarction Study, patients with SDNN (standard deviation of all normal RR intervals) of <50 ms obtained from predischarge 24‐hour Holter recordings had increased 1‐year mortality (independent of LVEF) compared to patients with SDNN between 50 and 100 ms and those with SDNN >100 ms.34 The use of HRV to predict SCD risk in other patients with CAD is less well established and partly complicated by the effects of ischemia and balloon treatment on HRV indices.35, 36 In addition, HRV is influenced by a variety of variables such as age, gender, and medication (eg, thrombolysis, antiarrhythmic drugs, β‐blockers and angiotensin‐converting enzyme inhibitors)32, 37, 38 and cannot be evaluated in patients with atrial fibrillation or frequent arrhythmias. Thus, the use of HRV alone for risk stratification of SCD is limited.

Heart Rate Turbulence

Heart rate turbulence (HRT) is a recently described ECG phenomenon that reflects the minute hemodynamic disturbance caused by a single PVC, and in doing so provides information on cardiovascular autonomic function. An example of the HRT curve obtained from a single PVC is shown in Figure 2. HRT parameters, turbulence onset, and turbulence slope, are affected by a number of variables, including age, certain medication, left ventricular function, and coronary revascularization.39 The clinical use of HRT has been most extensively studied for risk prediction of SCD in patients following myocardial infarction. Several large‐scale retrospective and prospective studies have provided strong evidence that HRT is a powerful independent predictor of SCD risk.29, 40, 41 The Noninvasive Risk Assessment Early After a Myocardial Infarction study investigators performed a number of autonomic function tests, including measurements of HRT, in 322 patients with LVEF <50% after myocardial infarction and demonstrated that these tests could reliably identify those at high risk of serious cardiac events.41 Another recent prospective study involving 2343 patients following myocardial infarction found that the combination of HRT and deceleration capacity could be used together to identify a high‐risk group equivalent in size and mortality to patients with LVEF <30%.42 In contrast, in a retrospective analysis of 884 patients enrolled in the Multicenter Autonomic Defibrillator Implantation Trial II study, HRT parameters were not found to be predictive of outcome, after adjustment for confounding covariates.43 However, HRT parameters in this study were obtained from 10‐minute ECG recordings, which may have limited the ability to derive accurate HRT data.

Figure 2.

Screen shot of heart rate turbulence (HRT) measurement from a single premature ventricular complex (PVC). The HRT curve shown (A) is calculated from the single PVC (B). Similar HRT curves can be produced from an average of all the PVCs detected during a 24‐hour Holter recording. The areas of early acceleration and late deceleration are annotated. HRT is quantified by 2 parameters: turbulence onset (TO) and turbulence slope (TS). TO (expressed as a percentage) is the relative change of RR intervals from before to after the PVC. TS is the slope of the steepest regression line fitted over the sequences of 5 consecutive sinus rhythm RR intervals within the 15 RR intervals after the PVC (expressed in ms/RR interval).

Whether HRT parameters are useful in predicting the risk of SCD among patients with stable CAD and no history of myocardial infarction remains to be determined in large‐scale prospective studies. Apart from the limitations of multiple variables that can affect HRT, other considerations include the current need to obtain 24‐hour Holter recordings and the requirement for the presence of PVCs. Future strategies may include shorter recording periods with or without pacing strategies to evoke PVCs and obtain HRT measurements.

T‐Wave Alternans

Electrical alternans of the T wave (ie, alternating amplitude from beat to beat) on the ECG is thought to be due to dispersion of repolarization and has been demonstrated to be associated with life‐threatening ventricular arrhythmias. T‐wave alternans can be measured during exercise (microvolt T‐wave alternans [MTWA]) or during ambulatory Holter‐based recordings (modified moving average analysis).44, 45 Figure 3 shows representative traces of MTWA obtained during a 24‐hour Holter recording in a patient with left ventricular dysfunction. Measurements of T‐wave alternans may be predictive of arrhythmic events and mortality in patients with CAD. For example, Holter analysis of the 49 (out of 1649) patients in the Cardiovascular Health Study with usable Holter recordings who suffered SCD demonstrated that increased T‐wave alternans was associated with SCD.31 In a prospective study involving 1003 patients with CAD (322 test cohort and 681 validation cohort), the modified moving average method of assessing T‐wave alternans was found to be a strong, independent predictor of risk (that was equivalent to impaired ejection fraction) during 48 months of follow‐up.46 MTWA has also been demonstrated to be a powerful predictor of life‐threatening arrhythmias and SCD in patients following myocardial infarction, both with and without depressed LV function.47, 48, 49 In a recent prospective multicenter study involving 575 patients, Chow et al found that MTWA testing in patients with ischemic heart disease and LVEF <30% who already qualified for an ICD did not predict subsequent ventricular arrhythmic events, although MTWA non‐negative patients (ie, positive and indeterminate MTWA results) had significantly higher mortality compared with MTWA‐negative patients.50 The value of MTWA in risk stratification may actually be in deciding which patients are least likely to benefit from ICD insertion, as suggested by the Alternans Before Cardioverter Defibillator trial.51 This prospective, multicenter study was the first to use MTWA to guide prophylactic ICD insertion. The investigators demonstrated that MTWA achieved 1‐year positive and negative predictive values of 9% and 95%, respectively, and that its use in risk stratification was comparable to invasive electrophysiological study at 1 year and complementary when applied in combination.

Figure 3.

Abnormal microvolt T‐wave alternans in a patient with left ventricular dysfunction obtained from 24‐hour Holter recording. Traces of maximum and minimum T wave amplitudes (TWA) are superimposed for leads V5 and V1.

Conclusion

SCD is not a single entity but rather is related to a variety of factors including coronary anatomy, degree of myocardial ischemia, left ventricular dysfunction, presence of scar/arrhythmic substrate, and genomic factors. Consequently, it is unlikely that any single test will be able to adequately risk stratify which patients with CAD are at greatest risk of SCD and warrant insertion of an ICD for primary prevention. The 12‐lead ECG parameters and ECG‐based tests described in this review provide useful information on the presence of underlying arrhythmic substrate and SCD risk. The advantages and disadvantages of each of these tests are summarized in the Table 1. As can be seen from the last column in this table, the current practical value of many of these noninvasive tests for risk stratification and selection of ICD recipients are limited, especially if used alone. Nevertheless, some of these tests may be useful to clinicians who are responsible for risk stratification of patients with CAD to help guide patient discussion. The decision to insert an ICD for primary prevention is never an easy one; several factors need to be considered (eg, patient preferences, comorbidities, cardiac factors) and weighed against the potential adverse effects of ICDs (eg, device‐related complications, inappropriate shocks). Thus, more objective, noninvasive measures of risk can be useful in some situations. However, it should also be noted that most of these tests are not featured in current guidelines for selection of ICD recipients, which are based mainly on the results of prospective multicenter trials. Therefore, as with many other aspects of modern medicine, clinicians need to find the right balance between clinician‐patient dialogue, evidence‐based medicine, and the adherence to current guidelines.

Table 1.

Summary of Electrocardiograph‐Based Tests That Provide Information on the Risk of Sudden Cardiac Death in Patients With Coronary Artery Disease

ECG‐Based Test	Advantages	Disadvantages/Limitations	Practical Value for Risk Stratification
12‐lead ECG	Cheap, quick and easy to perform; can be obtained serially at each follow‐up visit to reassess risk; large databases can be generated and analyzed retrospectively/ prospectively	Many abnormal parameters are markers of increased mortality, rather than specifically SCD; low positive predictive and negative predictive accuracies; subject to interobserver variability (unless automated software is used); considerable overlap in some parameters between healthy subjects and patients	Remains a standard investigation in patients with CAD; low positive and negative predictive accuracies for SCD limit its practical use in risk stratification and selection of ICD candidates
Signal‐averaged ECG	Easy and quick to perform; high negative predictive accuracy; can be used in patients with AF	Low positive predictive accuracy; numerous negative studies, especially in current era of interventional cardiology; better at predicting risk of VT than VF; normal standards for patients with bundle branch block or paced rhythm have not been established	Improved risk stratification when used in combination with other tests; probably more useful in identifying low‐risk patients; not useful alone for risk stratification
Standard 24‐hr Holter	Provides information on other arrhythmias in patients with CAD (eg, AF, heart block); standard test, easy to perform; can be used in patients in AF or paced rhythms	Low sensitivity and specificity	Most promising use is in combination with other parameters (eg, HRV and HRT) obtained from Holter recordings; not useful alone for risk stratification
Heart rate variability	Can be automatically recorded with standard Holter (using additional software); short (2–30 min) and longer (24 h) measurements are possible	Cannot be reliably assessed in patients with AF or frequent PVCs Influenced by a number of factors (eg, age, medication); may be affected by functional state of sinus node; short‐term measurements in risk prediction have not been well tested; no consensus on which parameters of HRV or method of assessment is best	Current practical use for risk stratification is limited as a consensus opinion on which parameters of HRV to record and which method of assessing HRV is required
Heart rate turbulence	Value in risk prediction post‐AMI supported by several recent large‐scale prospective studies; provides prognostic information in patients with normal and impaired LVEF	Optimal time post‐AMI to perform the test has not been established; can only be performed in patients in SR with a significant number of PVCs; use in patients with CAD and no history of AMI not well established	A promising test for risk prediction that can be used with other Holter‐based measurements; limited practical use at present in the absence of clear guidelines for risk stratification
T‐wave alternans	Easy to perform; can use existing equipment or modification of equipment; high negative predictive accuracy	Can only be used in patients in SR; “clean” ECG trace required (difficult to obtain during exercise); indeterminate result if target heart rate not achieved during exercise; low positive predictive accuracy	Useful in risk stratifying patients with impaired and preserved LVEF; useful role in determining which patients are unlikely to benefit from ICD insertion; improved risk stratification when used in combination with other tests; clear guidelines awaited for practical use for risk stratification

Abbreviations: AF, atrial fibrillation; AMI, acute myocardial infarction; CAD, coronary artery disease; ECG, electrocardiogram; HRT, heart rate turbulence; HRV, heart rate variability; ICD, implantable cardioverter defibrillator; LVEF, left ventricular ejection fraction; PVC, premature ventricular complex; SCD, sudden cardiac death; SR, sinus rhythm; VF, ventricular fibrillation; VT, ventricular tachycardia.

The future of SCD risk prediction in patients with CAD probably lies in the development of multivariate risk models that integrate clinical parameters with more specific tests of arrhythmic risk. These tests may be invasive or noninvasive in nature and could include some of the ECG‐based tests described in this review. The importance of such models is their ability to assess the risk of SCD in CAD patients with relatively preserved LVEF. This group represents the majority of patients with CAD and forms the majority of SCD victims, although they are not included in current guidelines for the selection of prophylactic ICD candidates. Such models require prospective validation before they can be incorporated into clinical guidelines. Another potentially even greater challenge to the cardiology community is in the identification of individuals with silent CAD in whom the first manifestation of disease is cardiac arrest or SCD. Genomic markers for arrhythmic susceptibility are likely to play an important role in these individuals and may form the basis of refined risk stratification for SCD, with or without ECG‐based tests.