 |
|
|
Signal-averaged electrocardiography (SAECG) is a technique involving computerized analysis of segments of a standard surface electrocardiogram. It is used for detecting small electrical impulses, termed ventricular late potentials, that follow the QRS segment. They are embedded in the electrocardiogram but ordinarily obscured by skeletal muscle activity and other extraneous sources of "noise" encountered in recording a standard electrocardiogram. Ventricular late potentials in patients with cardiac abnormalities, especially coronary artery disease or following an acute myocardial infarction, are associated with an increased risk of ventricular tachyarrhythmias and sudden cardiac death. Proponents of SAECG claim that it can obviate the need for invasive techniques commonly used to identify high-risk patients for interventions that treat or prevent ventricular tachyarrhythmia and sudden death.
No randomized clinical trials evaluating SAECG have been completed; data from an ongoing National Institutes of Health-sponsored clinical trial are expected to be available in 3-4 years. The current data on SAECG show relatively consistent high negative predictive values, poor positive predictive values, and variable sensitivity and specificity when the technique is used on patients with cardiomyopathy or following a myocardial infarction. The available evidence also indicates that combining SAECG with other tests of cardiac function is superior to using any single test for risk. The utility of SAECG alone as an indicator of risk remains to be proven. SAECG combined with other standard tests of risk has been demonstrated to have clinical utility in patients following an acute myocardial infarction. Other patient populations have not been conclusively shown to benefit from its use.
The Center for Practice and Technology Assessment (CPTA) evaluates the risks, benefits, and clinical effectiveness of new or established medical technologies. In most instances, assessments address technologies that are being reviewed for purposes of coverage by federally funded health programs.
The CPTA assessment process includes a comprehensive review of the medical literature and emphasizes broad and open participation from within and outside the Federal Government. A range of expert advice is obtained by widely publicizing the plans for conducting the assessment through publication of an announcement in the Federal Register and solicitation of input from Federal agencies, medical specialty societies, insurers, and manufacturers. The involvement of these experts helps ensure inclusion of the experienced and varying viewpoints needed to round out the data derived from individual scientific studies in the medical literature.
The CPTA staff analyzed and synthesized data and information received from experts and the scientific literature. The results are reported in this assessment. Each assessment represents a detailed analysis of the risks, clinical effectiveness, and uses of new or unestablished medical technologies. If an assessment has been prepared to form the basis for a coverage decision by a federally financed health care program, it serves as the Public Health Service's recommendation to that program and is disseminated widely.
The CPTA is one component of the Agency for Health Care Policy and Research (AHCPR), Public Health Service, Department of Health and Human Services.
Douglas B. Kamerow, M.D., M.P.H., Director, Center for Practice and Technology
Assessment
John M. Eisenberg, M.D., Administrator, Agency for Health Care
Policy and Research
Questions regarding this assessment should be directed to:
Center for Practice and Technology Assessment
Agency for Health Care Policy and Research
6010 Executive Boulevard, Room 316
Rockville, MD 20852
Telephone: (301) 594-4026
Acc:: Accuracy
ART:: Arrhythmia research technology
BBB:: Bundle branch block
CAD:: Coronary artery disease
CAST:: Cardiac arrhythmia suppression trials
CFR:: Case fatality rate
CHF:: Congestive heart failure
CI:: Confidence interval
dQRS:: Duration of the filtered QRS complex >106 ms
ECG:: Electrocardiogram
FD:: Frequency domain
FDA:: Food and Drug Administration
HCFA:: Health Care Financing Administration
IDC:: Idiopathic dilated cardiomyopathy
LAS:: Low amplitude signal
LVEF:: Left ventricular ejection fraction
LVH:: Left ventricular hypertrophy
MeSH:: Medical subject heading
MI:: Myocardial infarction
MUGA:: Multigated acquisition
NPV:: Negative predictive value
NS:: Not stated
NSR:: Normal sinus rhythm
NYHA:: New York Heart Association
PES:: Programmed electrical stimulation
PPV:: Positive predictive value
RMS:: Root mean square voltage during the last 40 ms of the QRS complex <25 µV
RNV:: Radionuclide ventriculography
SAECG:: Signal-averaged electrocardiography
SD:: Sudden death
1SD:: One standard deviation
Se:: Sensitivity
Sp:: Specificity
STA:: Spectral turbulence analysis
TD:: Time domain
VAT:: Ventricular activation time
VF:: Ventricular fibrillation
VLP:: Ventricular late potentials
VT:: Ventricular tachyarrhythmia
The Center for Practice and Technology Assessment, Agency for Health Care Policy and Research, conducted this technology assessment of the safety, effectiveness, and clinical utility of signal-averaged electrocardiography (SAECG) at the request of the Health Care Financing Administration. Biomedical signal-averaging techniques have been used in neuroscience to study cortical sensory-evoked potentials, in obstetrics to separate fetal and maternal electrocardiograms, and in cardiology to detect ventricular late potentials (VLPs), primarily in patients with ischemic heart disease.(1, 2) During the past 2 decades, proponents have attempted to find a reliable use for SAECG in evaluating patients with sustained or nonsustained ventricular tachyarrhythmia after myocardial infarction (MI), nonneurogenic syncope, and cardiomyopathy.(3, 4) Other proposed uses of SAECG include monitoring heart transplant rejection and antiarrhythmic drug therapy. This assessment evaluates the ability of SAECG to predict sudden death or ventricular tachyarrhythmia in the presence of VLP.
A literature search was performed using the online electronic retrieval systems PaperChase and MEDLINE, with the following keywords or medical subject heading (MeSH) terms: VLPs and electrocardiography (ECG), and spectral analysis (limited by signal averaging). Approximately 500 articles, published in the English language between 1985 and September 1996, were retrieved. Citations were added from reference lists of appropriate articles and textbooks. Greater weight was given to data from prospective studies in which a control or reference group was compared with the study population.
Approximately 1.5 million Americans suffer myocardial infarctions each year.(5) Of these, approximately one half are hospitalized, and one third die before reaching the hospital.(4-6) Of those who reach the hospital, 85 percent experience residual ventricular irritability(7-9) in the form of primary ventricular fibrillation (4-8 percent),(10) ventricular tachyarrhythmia (6-10 percent),(11) or sudden death (5 percent).(12) Only 10 percent of all hospitalized acute MI patients undergo full electrophysiologic testing (20 percent of whom receive automatic implantable cardioverter defibrillators).(13) Although the case fatality rate for ischemic heart disease has declined by 50 percent over the past 30 years (partially attributed to coronary intensive care units and early thrombolytic and antiarrhythmic drug interventions), 50-60 percent of patients with ischemic heart disease die suddenly.(14) The case fatality rate for patients hospitalized with acute MI, complicated by congestive heart failure or primary ventricular fibrillation, is 48 percent, vs. 1.5 percent in uncomplicated acute MI patients (p = 0.001).(10) Over the past 15 years, the incidence of primary ventricular fibrillation during an acute MI has remained unchanged (4-8 percent/year). Risk factors associated with repeat infarct or sudden death are recurrent myocardial ischemia, moderate to severe reduction in the left ventricular ejection fraction, and spontaneous ventricular ectopy (>10/hour).(15, 16) The risk factors for ventricular tachyarrhythmia are age, congestive heart failure complicating an acute MI, and prior MI.(9, 17) The 1-year case fatality rate for patients with a second MI or sustained chronic angina is twice that of persons with an initial acute MI.
Sudden cardiac death results from malignant tachyarrhythmias (ventricular tachycardia or fibrillation).(18) Most patients who die suddenly from cardiac causes have underlying severe coronary artery disease, infiltrative or hypertrophic cardiomyopathies, or primary electrical disturbances of the cardiac conduction system.(19) A cardiac etiology, e.g., sinus node disease, ventricular tachyarrhythmia, supraventricular tachycardia, heart block, or syncopal episodes, can be verified in only a small proportion of patients (mean 14 percent); the etiologies of most cases remain undetermined.(3)
The two principal mechanisms responsible for ventricular tachyarrhythmia are reentry and enhanced automaticity.(14) Spontaneous, sustained ventricular tachyarrhythmia is thought to originate from the reentry of aberrant, asynchronous electrical impulses in the <25 µV range, called ventricular late potentials, caused by reduced conduction velocity and/or elongated inhomogeneous conduction pathways.(1, 20, 21) Ventricular late potentials are generated from viable isolated cardiac muscle bundles bordering an infarction or ischemic zone of the myocardium and are propagated through a zone of slow conduction (arrhythmogenic substrate) to a zone of healthier myocardium where repolarization has already begun. Enhanced automaticity leads to repetitive firing from a single focus within the heart and may be precipitated by hypoxia, ischemia, acid-base disturbances, and drugs. Late potentials are believed to be precursors of sudden death, ventricular tachyarrhythmia, and/or ventricular fibrillation.(20)
Ventricular late potentials were first described in the late 19th century and consist of small electrical impulses (1-2 µV) that are embedded within the electrocardiographic junction between the electrocardiograph wave complex or interval complex and ST segment (QRS-ST) of an ECG.(21) They are obscured by skeletal muscle activity and other extraneous sources of noise within the 5- to 20-µV range.
In natural history studies of at-risk populations, the prevalence of VLP is higher in patients with a history of coronary artery disease and MI (21-43 percent), spontaneous and induced sustained ventricular arrhythmias (primarily ventricular tachycardia and ventricular fibrillation), and left ventricular aneurysms than in patients with normal ventricular function.(22, 23) Ventricular late potentials rarely occur without myocardial disease (0-6 percent of studied control populations).(1, 24)
Signal-averaging techniques, which reduce the noise (low-frequency, high-amplitude signals) interfering with the surface ECG, have been used since the 1970s to detect VLP.(24-26) A positive SAECG is one in which VLP is detected. The SAECG procedure involves recording the-surface ECG during a period of normal sinus rhythm, followed by amplification of the entire ECG (103 - 108 x the original signal), which exaggerates all electrical signals. The principal noise removed by signal averaging is from skeletal muscle. Once amplified, the voltage signal from each lead undergoes analog to digital conversion at frequent fixed time intervals. A computer signal-averages the signal from each lead against the midpoint of the R wave or the peak slope in front of the QRS waves in the normal sinus rhythm template, previously chosen by the operator or assigned by the software. More than 400 beats are recommended for averaging, although in older systems averages were calculated with fewer than 250 beats. An abnormal tracing from a patient with spontaneous sustained ventricular tachycardia shows low-amplitude signal in the terminal QRS complex, which extends 210 ms after QRS onset. Mean voltage (V40) is in the terminal 40 ms of the QRS.
Currently, there are two methods of body-surface ECG signal-averaging: time-domain (or ensemble averaging or high pass filtering) and frequency-domain (or spatial or spectral averaging). Time-domain averaging is a computer-driven sequential digital process based on the vector analysis of three orthogonal ECG leads (X, Y, and Z) that separates and extracts high-frequency late potentials from lower frequency ST segments by the use of directional filters, enabling filter output to correspond in time to signal input. Most of the noise is eliminated by electrically isolated amplifiers, shielded lead systems, and filters to reduce aberrant noise to <1 µV.
Conventional time domain analyses have limitations that to some degree can be mitigated by other more detailed analyses, including frequency domain analyses.(27, 28) This involves further filtering of overlapping QRS segments, subjecting them to fast-Fourier transformation, and plotting the analysis in a three-dimensional representation.
No consensus exists on the definition of VLP in time-domain or frequency-domain analysis.(24, 28) Investigators have used increases in spectral amplitude >30 dB) at frequencies between 40 and 200 Hz and prolonged ventricular activation times to show an abnormal frequency domain. Time-domain parameters (root mean square voltage during the last 40 ms of the QRS complex <25 µV [RMS40], duration of low-amplitude signals during the last 40 ms >38 ms [LAS40], or duration of the filtered QRS complex >106 ms [dQRS]) are used to mark the presence of VLP in the SAECG.(30) Ventricular late potentials are defined relative to the beginning and endpoints of the filtered QRS, which are directly influenced by residual noise in the system after data acquisition and the average number of beats. Therefore, many opportunities exist for misclassification among the different SAECG systems.(31)
Age, gender, thrombolytic therapy, antiarrhythmic drug use,(32, 33) timing of the SAECG in the postinfarction period,(34) and the period from diagnosis of ventricular tachyarrhythmia to SAECG testing(35) are nonsystem factors that may influence the incidence of the VLP. McGuire et al(36) found no difference in age, gender, site of infarct or thrombolytic therapy in SAECG-positive or SAECG-negative patients. Malik(37) confirmed that there was no difference in the incidence of positive SAECG in patients receiving thrombolytic therapy. A more recent paper, however, concluded that SAECG was abnormal more often when thrombolytic therapy was omitted.(38)
There are recording difficulties in both time-domain and frequency-domain systems. Time-domain is neither suitable for patients with QRS complexes >120 ms nor able to detect changes in late potential activity. Frequency domain shows significant beat-to-beat variability in morphology, duration, and latency. The ability to detect this variability, however, remains investigational. Some investigators believe that these problems underlie the biologic variation of the myocardium (tachyarrhythmias, myocardial oxygen uptake) or the influence of drugs during ischemia rather than poor reproducibility of the method.(29)
In 1991, the American Heart Association, European Heart Association, and American College of Cardiology Task Force Committee on SAECG noted in a joint scientific statement that although commercially available SAECG devices had different technical systems, algorithms, and recording units, and a definition of a positive SAECG could not be agreed on, standards should be established for data acquisition and analysis, and the role of SAECG in clinical decisionmaking should be defined.(1) At that time, no consensus had been reached on technical design considerations because of ongoing component development. No particular method of data analysis was endorsed. The committee pointed out that multivariate analysis showed an independent effect of a positive SAECG from other risk markers such as reduced left ventricular ejection fraction or the presence of sustained ventricular tachyarrhythmias within 1 year of an acute MI. Earlier investigations showed a positive correlation between VLP and inducibility of ventricular tachyarrhythmia during electrophysiologic testing.(39) Proponents of noninvasive SAECG believed it could stratify patients thought to be at risk for sudden death or ventricular tachyarrhythmia after acute MI without invasive procedures such as coronary angiography or programmed electrical stimulation, particularly in light of its high negative predictive value. However, as the committee noted, "the relatively low positive predictive value of SAECG in post-MI patients emphasizes the need for continued methodological refinements that will increase its diagnostic power."(1)
Since 1991, little standardization of SAECG components has taken place. A multicentered clinical trial sponsored by the National Heart, Lung, and Blood Institute is underway as a spinoff of the cardiac arrhythmia suppression trials (CAST) to help evaluate the role of SAECG in the diagnosis of ischemic heart disease. The results of this trial should be available in 3-4 years.
In January 1996, the American College of Cardiology published an expert consensus document on SAECG with the following findings:(40)
Established value of SAECG:
Stratification of risk of developing sustained ventricular arrhythmias in patients recovering from MI who are in sinus rhythm without electrocardiographic evidence of bundle branch block (BBB) or intraventricular conduction delay (QRS complex >120 ms).
Identification of patients with ischemic heart disease and unexplained syncope who are likely to have inducible sustained ventricular tachycardia.
Valuable in clinical care, further supportive evidence desirable:
Stratification of risk of developing sustained ventricular arrhythmias in patients with nonischemic cardiomyopathy.
Assessment of success of operation for sustained ventricular tachycardia.
Promising but currently unproved:
Detection of acute rejection of heart transplants.
Assessment of efficacy or proarrhythmic effects of antiarrhythmic drug therapy in patients with ventricular arrhythmias.
Assessment of success of pharmacologic, mechanical, or surgical interventions to restore coronary artery blood flow.
Not indicated:
Patients with ischemic heart disease and documented sustained ventricular arrhythmias.
Stratification of risk of developing sustained ventricular arrhythmias in asymptomatic patients without detectable heart disease.
| Manufacturer | Model No. |
| Mortara Instrument | 101 HiRes |
| Marquette Electronics | MAC I-302 MAC 12 Hi-Res |
| Arrhythmia Research Technology | 1200 EPX |
| Aerotel | Hi-PEC |
| Nicolet Instrumentation | MED-80 |
| Fidelity Medical | LP-3 LP 3000 LP Plus |
| Corozonix | Predictor Predictor II Predictor IIc |
| Del Mar Avionics | Micropotential Analyzer 705A |
| Wex Medical Instrumentation | Multidomain Cardiac System HD-3L |
| Diagnostic Medical Instruments | Altair Hi-Res Recorder |
| Hewlett Packard | M1754A |
Published clinical data were reviewed to determine the evidence base for using SAECG in high-risk subpopulations. The literature, published within the past 5 years, has focused on the prognostic ability of SAECG in acute MI and cardiomyopathy, sensitivity analyses of SAECG components such as filters, and other spectral applications of time-domain data. The clinical data in this assessment address acute MI and cardiomyopathy, although the range of published studies addresses SAECG in the diagnosis of syncope, heart rejection, and antiarrhythmic drug efficacy.
| Preference | No. of patients | Device | Parameters defining VLPdQRSaRMS40bLAS40c | Definition of +SAECG | Se (%) | Sp (%) | +PV (%) | -PV (%) | Mean followup days (range) | End-pointsd (%) | No. +SAECG (%) during followup | ||
| Ahuja(41) | 262 | ART 1200 EPX | + | RMS40 | NS | NS | .15 | .98 | 315±74 | 17/262 (7) | 36/262 (14) | ||
| Breithardt(20) | 132 | Princeton signal average No. 4202 | + | LAS40 | .66 | .60 | .23 | .90 | 450±330 | 9/132 (7) | 59/132 (45) | ||
| Brembilla-Perrot(42) | 328 | LP 3000 (Cardionics) | + | + | Any 1 parameter | NS | NS | NS | NS | 1,034±794 | 16/328 (5) | 151/328 (46) | |
| Cripps(43) | 176 | ART 101 | + | + | + | Any 1 parameter | .82 | .81 | .22 | .99 | 450±210 (90-720) | 11/176 (6) | 41/176 (23) |
| de Chillou(44) | 114 | MAC 12/15 | + | + | + | Any 2 parameters | NS | NS | NS | NS | 330±90 (180-540) | 25/114 (22) | 26/114 (23) |
| Denes(45) | 787 | ART 1200 or Corozonix Predictor | + | + | + | dQRS | NS | NS | NS | NS | 300±90 | 92/787 (12) | 97/787 (12) |
| Denniss(46) | 306 | DEC 11/34 computer (Fortran (program) | + | dQRS | .66 | .78 | .24 | .96 | 360 | 43/306 (14) | 80/306 (26) | ||
| Do(47) | 90 | Corozonix Predictor | + | + | + | All 3 parameters | 1.0 | .78 | .16 | 1.0 | NS | NS | 26/90 (29) |
| El-Sherif(34) | 156 | ART 1200 EPX | + | + | dQRS, or RMS40 | .35 | .79 | .38 | .76 | (0-365) | 20/156 (13) | 81/156 (52) | |
| El Sherif(48) | 1,158 | ART 1200 EXP or Corozonix Predictor | + | + | + | dQRS, or RMS40 | .48 | .90 | .17 | .98 | 310±96 (0-365) | 45/114 (40) | 139/158 (12) |
| Farrell(49) | 416 | ART 1200 EPX | + | + | + | Any 2 parameters | .63 | .81 | .17 | .81 | 612 (1-1,112) | 47/416 (11) | 89/418 (21) |
| Gomes(50) | 115 | ART 101 | + | + | + | Any 1 parameter | .63 | .84 | .38 | .93 | 420±240 | 16/115 (14) | 51/115 (44) |
| Gomes(51) | 102 | ART 101 | + | + | + | Any 1 parameter | .75 | .60 | .24 | .94 | 360±180 | 15/102 (15) | 30/102 (29) |
| Hohnloser(52) | 173 | NS | + | + | + | Any 2 parameters | .56 | .78 | .12 | .97 | 360±150 | 9/173 (5) | 41/173 (24) |
| Kuchar(53) | 210 | ART | + | + | LAS40 or dQRS | NS | .65 | .17 | NS | 220 (median) (180-720) | 15/210 (7) | 78/210 (37) | |
| Mäkijärvi(54) | 130 | Hewlett Packard HP1000 | + | + | + | Variable | .81 | .79 | .44 | .95 | 720 | NS | NS |
| Malik(55) | 553 | ART 1200 EPX | + | + | + | Any 2 parameters | .58 | .72 | .09 | NS | 360 | 50/553 (9) | 50/553 (9) |
| McClements(56) | 301 | Corozonix Predictor | + | + | + | Any 1 parameter | .64 | .80 | .11 | .98 | 367 | 13/301 (4) | 123/301 (4) |
| McGuire(36) | 50 | ART 101 | + | + | dQRS or RMS40 | .80 | .47 | .27 | .90 | 10 | 10/50 (20) | 29/50 (58) | |
| Mercando(57) | 121 | ART 1200 EPX | + | + | + | Any 2 parameters | .52 | .68 | .32 | .83 | 870±420 | 48/121 (40) | 26/121 (21) |
| Pedretti(58) | 303 | Fidelity Medical LP 3000 | + | + | + | Any 2 parameters | .79 | .75 | .17 | .98 | 450±210 | 19/303 (6) | 77/303 (25) |
| Reinhardt(59) | 769 | NS | + | dQRS | .76 | .63 | .10 | .98 | 180 | 33/778 (4) | 266/778 (29) | ||
| Richards(60) | 225 | DEC 11/34 | + | dQRS | .55 | .82 | .17 | .97 | 740 (median) | 34/225 (15) | NS | ||
| Rodriguez(61) | 190 | NS | + | + | dQRS | .39 | .91 | .48 | .88 | 750 | 33/190 (17) | 23/178 (12) | |
| Steinberg(62) | 182 | Corozonix Predictor | + | + | + | Any 1 parameter | .69 | .64 | .15 | .95 | 420±210 (30-960) | 38/182 (21) | 71/182 (39) |
| Turitto(27) | 90 | ART 101 PCD | + | + | dQRS or RMS40 | .75 | .78 | .14 | .98 | 0-5 | 1 (1) | 12/89 (13) | |
| Verzoni(63) | 220 | ART 101 | + | + | + | Any 2 parameters | .83 | .73 | .08 | .99 | 360 (45-360) | 62/220 (28) | 46/192 (24) |
| Zimmerman(64) | 223 | Siemens Mingograph 82 | + | + | LAS40 +dQRS | .67 | .80 | .16 | .98 | 450±135 | 17/223 (8) | 45/223 (20) | |
Duration of the filtered QRS complex > 106 ms.
Root mean square voltage during the last 40 ms of the QRS complex <25 µV.
Duration of low-amplitude signals during the last 40 ms >38 ms.
SD, VT, or MI.
Abbreviations: ART = arrhythmia research technology; NS = not stated; +PV = positive predictive value; -PV = negative predictive value; Se = sensitivity; Sp = specificity; SD = sudden death; VT = ventricular tachycardia; MI = myocardial infarction.
Ahuja et al(41) prospectively studied 262 patients with an acute MI. During the following 10.5 ± 2.4 months, 36 patients (14 percent) had a positive SAECG and 17 patients (7 percent) had arrhythmic events. The best predictive results were seen in patients with an anterior MI.
Breithardt et al,(20) in an early prospective study designed to identify patients at risk for ventricular tachyarrhythmias, found VLPs in 45 percent of patients. They pooled these data with results from seven earlier prospective studies (2,039 patients) showing either acute MI or chronic coronary artery disease and found that VLPs were present in 13.9 percent of patients with coronary artery disease who were followed for 20 ± 5 months and in 4-40 percent of patients with acute MI followed for 39 ± 15 months. In patients with and without VLP, sudden death (<1 hour after MI) occurred in 3.6-21 percent and 0.9-4 percent of patients; death from cardiac causes was 7-40 percent and 0-5 percent, respectively; and the occurrence of ventricular tachyarrhythmia was 2-29 percent and 1-5 percent, respectively.
Brembilla-Perrot et al(42) performed SAECG on 328 patients without bundle branch block. During the relatively long followup period of 3.7 ± 2.2 years, 46 percent of patients had a positive SAECG, but only 5 percent died suddenly or suffered from ventricular tachyarrhythmia.
Cripps et al(43) prospectively studied 176 patients 7 days post-MI. Multiple regression analyses of seven variables indicated exercise testing as the only independent variable reliably predicting ischemic events. However, SAECG did provide prognostic information concerning arrhythmic events (sensitivity 82 percent).
De Chillou et al(44) prospectively studied 114 patients post-MI with SAECG performed within 48 hours of infarct (early) and again 6-18 months (late) after hospital discharge. Results indicated that a change from an early normal to a late abnormal SAECG was highly indicative of recurrent ischemic events.
Denes et al(45) prospectively studied SAECG after thrombolytic therapy and/or angioplasty during the early phase of acute MI in 787 patients. An abnormal SAECG was found to be predictive of an increased incidence of arrhythmias in all patients independent of prior thrombolytic therapy and/or angiography.
Denniss et al(46) studied a subset of patients with recent transmural MI without bundle branch block (n = 306) to determine the incidence of inducible ventricular tachyarrhythmia by programmed electrical stimulation and the prevalence of VLP (delayed ventricular activation times >140 ms after QRS onset) in post-MI patients.(44) Using frequency-domain SAECG, they found that 26 percent of patients with recent MI had VLPs detected by SAECG.
Do et al(47) in a retrospective study of 90 patients used time-frequency distributions to detect late potentials. Their results from the binomial transform were comparable to those from the standard time-domain criteria.
El-Sherif et al(34) conducted serial SAECG recordings in 156 patients up to 60 days after an acute MI and found that abnormal recordings during days 6-30 had the most significant relation to arrhythmias during the first year after an infarct.(32) They confirmed that early SAECG testing (within 5 days of an acute MI) or late development of an abnormality in the second month after infarction was not predictive of ventricular tachyarrhythmia in the post-MI period.
El-Sherif et al(48) also coordinated the recruitment of 1,211 post-MI patients at 10 medical centers. After excluding patients with bundle branch block, 1,158 patients were followed up for arrhythmic events for 1 year (10 ± 3 months) postinfarct. Through regression modeling with all clinical, Holter, and SAECG variables, the investigators found that the dQRS at 40 Hz at a cutpoint of >120 ms was the most significant predictor (p = 0.0001) of ventricular tachyarrhythmia. The positive predictive value, however, was only 17 percent. When combined with an abnormal Holter result and a left ventricular ejection fraction of <40 percent, the positive predictive value of dQRS rose to 32 percent. Negative predictive values of dQRS were consistently high for SAECG alone and improved when SAECG, Holter results, and left ventricular ejection fraction of <40 percent were added to the regression model.
Farrell et al(49) noted a 17 percent positive predictive value for SAECG while also showing that Killip Class II, heart rate variability, and left ventricular ejection fraction best predicted ventricular tachyarrhythmia when combined with SAECG.
Gomes et al(50) prospectively studied the prognostic significance of SAECG variables relative to other clinical variables and found that the predictive value of SAECG was equal to the left ventricular ejection fraction in patients with an inferior infarction and better than the left ventricular ejection fraction in patients with an anterior wall infarction. They calculated 5-month survival rates of 86 percent, and 72 percent in patients with ventricular tachyarrhythmia, <30 percent left ventricular ejection fraction, and positive SAECG, respectively.(48) In an earlier prospective study of SAECG in 102 patients, Gomes et al(51) concluded that the combination of an abnormal SAECG, an abnormal left ventricular ejection fraction, and the presence of high-grade ectopic activity was better in identifying high-risk patients than a single abnormal variable.
Hohnloser et al(52) prospectively studied 173 patients with an acute MI treated with thrombolysis. Late potentials were seen in 24 percent of patients. Multivariate analysis indicated that an occluded infarct-related artery and the presence of regional wall motion abnormalities were strong independent predictors for the development of VLPs.
Kuchar et al(53) performed SAECG during the hospitalization of 200 patients following an acute MI. Testing was performed 11 ± 6 days after initial presentation. Seventy-eight patients had an abnormal SAECG, a finding that identified 93 percent of patients with an arrhythmic event.
Mäkijärvi et al(54) studied 150 individuals: 26 patients with documented ventricular tachyarrhythmia (92 percent post-MI), 104 patients with acute MI without ventricular tachyarrhythmia, and 20 voluntary controls without evidence of heart disease. The purpose of this study was to determine the optimum filter settings that could be used to separate patients with and without ventricular tachyarrhythmia. All patients had decreased dQRS as the filter settings were increased from 25 Hz to 100 Hz, but the ventricular tachyarrhythmia group increased at all frequencies (p = 0.001). Sensitivity and specificity were simultaneously maximized (81 percent and 79 percent, respectively) by the dQRS parameter using a 25-Hz filter. The most sensitive single SAECG parameter was LAS50 using a 60-Hz filter. Logistic regression analysis demonstrated that the combined independent variables, dQRS, RMS50 using a 25-Hz filter, and LAS40 and RMS20 using a 80-Hz filter, were best at separating patients with ventricular tachyarrhythmia from those without ventricular tachyarrhythmia (X 2 = 33.4, p = 0.001).
Malik et al(55) recorded an SAECG in 553 survivors of acute MI before hospital discharge. SAECGs were analyzed using both time-domain and spectral turbulence analyses. Time-domain analysis was superior in predicting ventricular tachycardia/fibrillation during the first year of followup.
McClements and Adgey(56) found that ventricular tachyarrhythmia was best predicted by a combination of a positive SAECG, left ventricular ejection fraction (<40 percent), and the presence of congestive heart failure.
McGuire et al(36) performed SAECG every 48 hours in patients admitted to the hospital with an acute MI. The prevalence of a positive SAECG was 32 percent at presentation and increased progressively during hospital stay. A positive SAECG appeared to be useful in identifying patients at high risk of developing significant ventricular arrhythmias.
Mercando et al(57) performed SAECGs in 121 patients 6 months or more after an acute MI. All patients had complex ventricular arrhythmias. Forty-four patients had an abnormal SAECG, which alone was not predictive of mortality.
Pedretti et al(58) prospectively evaluated SAECGs in 303 survivors of acute MI. No single noninvasive prognostic variable was found valuable for identifying patients suitable for electrophysiologic studies. However, patients having both low left ventricular ejection fraction and a positive SAECG were considered eligible for programmed ventricular stimulation.
Reinhardt et al(59) evaluated the prognostic value of wavelet correlation functions of SAECGs in 769 men 2 to 3 weeks following an acute MI. The addition of wavelet correlation functions to SAECG analysis increased the prognostic value for significant arrhythmic events.
Richards et al(60) evaluated 225 patients 1-2 weeks following an acute MI and among the tests performed were electrophysiologic studies, left ventricular ejection fraction, SAECG, and Holter monitoring. The single best predictor of ventricular tachycardia or sudden death was inducible VT during electrophysiologic study.
Rodriguez et al(61) prospectively assessed 190 patients at varying intervals following a first acute MI. Sudden death was predicted only by left ventricular ejection fraction.
Steinberg et al(62) tested 182 consecutive patients after an acute MI. The risk associated with an abnormal SAECG was independent of left ventricular function and ventricular arrhythmias identified during Holter monitoring. Combined variables were found to be superior to single tests in identifying patients at high risk for VT or sudden death.
Turitto et al(27) found that the incidence of ventricular tachyarrhythmia during the first 2 months after an MI is not different in SAECG-positive and SAECG-negative patients.
Verzoni et al(63) prospectively studied the prognostic significance of SAECG in 220 survivors of an acute MI tested prior to hospital discharge and during the following year. SAECG was found to provide more prognostic information in identifying patients at risk for arrhythmic events than either Holter monitoring or left ventricular ejection function.
Zimmerman et al(64) assessed the effect of thrombolytic therapy on the incidence of ventricular late potentials in 223 consecutive patients surviving an acute MI. Incidence was indeed reduced, but its prognostic significance, however, required further study.
Recently, investigators, acknowledging the high number of false positives, have done sensitivity analyses of SAECG parameters to increase its sensitivity.(1, 2, 29-31, 37) However, operator-controlled variables and differences in SAECG component design do not permit an aggregated analysis because of variations in template matching, sampling frequency, filtering, baseline noise determinations, QRS onset and offset, the number of QRS complexes averaged, definitions of positive and negative SAECG, and the definitions of ventricular late potentials.
Others have reported positive predictive values ranging from 5-76 percent.(65) These wide ranges can be explained, in part, by variables in SAECG equipment and methodology. In contrast, the ability of SAECG to identify patients who remain free from sudden death or ventricular tachyarrhythmia during the post-MI period (up to 2 years) appears to be quite good (low false-negative rate and high negative predictive value). More than 10,000 MI patients have been studied using SAECG to identify patients at risk for sudden death or ventricular tachyarrhythmia.
In a review of 15 studies by the American College of Cardiology, the negative predictive value of SAECG had a range of 88-99 percent with a median of 97 percent.(40) The positive predictive value had a median of only 17 percent (range 8-48 percent).
The clinical utility of a negative SAECG obviates the need for additional testing to identify patients at high risk of sustained ventricular tachyarrhythmia after MI. A positive SAECG in post-MI patients would mandate further testing to determine the risk of sustained ventricular tachyarrhythmia and/or sudden death.
In prospective clinical studies published since the 1991 SAECG standards, the predictive ability of SAECG has not improved. The preponderance of evidence continues to support the clinical utility of SAECG combined with the left ventricular ejection fraction in stratifying post-MI patients with ventricular arrhythmias for low or significant risk of ventricular tachyarrhythmia or sudden death.(48-50, 53, 56, 58)
| Reference | Device | No. of patients | Mean followup days (range) | Frequency of endpoints (%)a | No. +SAECG (%) during followup |
| Keeling(67) | ART 1200 EPX | 64 | 540±420 (30-1,350) | 5/64 (8) | 22/64 (34) |
| Kulakowski(68) | ART 1200 EPX | 121 | 180-480 (median 330) | 9/121 (7) | 30/121 (24) |
| Mancini(4) | ART 1200 EPX | 86 | 300±150 | 9/86 (10) | 20/86 (23) |
| Middlekauff(69) | Corozonix | 62 | 218±154 | 9/62 (14) | 19/62 (31) |
| Silverman(70) | Corozonix Predictor II | 200 | 597±437 | 43/135 (32) | 61/135 (48) |
| Turitto(71) | ART 1200 EPX | 80 | 654±771 | 9/80 (11) | 12/80 (15) |
| Vester(72) | Corozonix Predictor I | 157 | 360±180 | 9/157 (6) | 37/57 (24) |
| Yi(28) | ART 1200 EPX | 84 | 720±540 (30-1,770) | 28/84 (33) | 31/84 (37) |
a Endpoints are SD (sudden death), VT (ventricular tachycardia), and MI (myocardial infarction).
Keeling et al(67) studied 64 consecutive patients with idiopathic, nonischemic, dilated cardiomyopathy with left ventricular hypertrophy, left ventricular ejection fractions of 28 ± 10 percent, and normal coronary arteries for 18 ± 14 months (range 1 to 45 months). They found positive predictive values of 88 percent and 57 percent in time-domain analysis (25- and 40-Hz filters, respectively) and 56 percent in spectral temporal mapping techniques in patients with any ventricular tachyarrhythmia. In patients with sustained ventricular tachyarrhythmia, the results were 38 percent, 21 percent, and 22 percent, respectively. Twenty-two patients had a positive SAECG during followup.
Kulakowski et al(68) studied 121 patients with hypertrophic cardiomyopathy who did not receive the antiarrhythmic drug, amiodarone. Patients were diagnosed by echocardiography, history, and physical examination. Both time- and frequency-domain SAECGs were done with 99 patients who had stopped taking antiarrhythmic medications. These patients were compared with a group of 44 age-matched controls. Seven percent of the patients had abnormal time-domain SAECGs. There were nine catastrophic events, eight of which occurred in men under the age of 30. Neither of the SAECG methods predicted these catastrophic events. The authors speculated that the mechanism for the sudden deaths was not ventricular tachyarrhythmia; thus, the poor positive predictive value of SAECG in cardiomyopathy.
Mancini et al(4) evaluated 86 patients with dilated nonischemic cardiomyopathy by time-domain SAECG. Patients were followed until death, sustained ventricular tachycardia (>120 beats per minute for >30 seconds), or a heart transplant. Twenty patients were SAECG-positive, and 66 patients had a normal SAECG. Mean followup time was 10 ± 5 months. Of the patients with left ventricular hypertrophy (39/86, 45 percent), most (36/39, 92 percent) had a normal SAECG. Patients with a normal SAECG had a 95-percent 1-year survival rate, and those with a positive SAECG had a 56 percent 1-year survival rate. Event-free (death, ventricular tachyarrhythmia, or urgent transplant) 1-year survival rate was 95 percent in the SAECG-negative group and 39 percent in the SAECG-positive group (p = 0.0001). Univariate analysis indicated abnormal SAECG, functional class, peak oxygen consumption, and history of ventricular tachyarrhythmia as the most significant variables (p = 0.0001 to p = 0.02). Multivariate predictors of any event were positive SAECG, functional class, cardiac index, standard QRS duration >120 ms, and age. As the authors of this study noted, although positive SAECG and functional class were independent predictors of death in patients with cardiomyopathy, left ventricular ejection fraction and peak exercise oxygen consumption were not (p = 0.62, p = 0.82, respectively). Explaining why these two parameters of left ventricular function were nonsignificant is difficult.
Middlekauff et al(69) concluded that the incidence of VLPs in a group of 62 patients depended on the etiology of the congestive heart failure. Late potentials were present in 31 percent of all patients and 40 percent of patients with evidence of an old MI but in only 3 percent of patients with nonischemic congestive heart failure.
Silverman et al(70) studied 200 patients with chronic congestive heart failure to investigate the relationship of heart failure to the results of the SAECG. Patient exclusions included permanent pacemaker or cardioverter/defibrillator insertion, or terminal illness. All patients were New York Heart Association (NYHA) Class II to IV (90 percent were Class III or IV). Cardiac catheterization or documented history of MI determined the diagnosis of ischemic cardiomyopathy, and all patients were prospectively stratified into normal SAECG (n = 74), abnormal SAECG (n = 61), or prolonged QRS (n = 65) categories. The investigators also divided the patients into ischemic (n = 88) or nonischemic (n = 112) groups. One of the few significant findings demonstrated that patients with nonischemic cardiomyopathy with a prolonged QRS have worse survival than the other two groups (p = 0.05). Patients with nonischemic disease with an abnormal SAECG had no worse a prognosis than patients with a normal SAECG. Patients with ischemic disease and an abnormal SAECG showed a trend toward a worse survival rate than that for those with a normal SAECG, but this was not statistically significant.
In an attempt to stratify for risk of arrhythmic events in a group of 80 patients with nonischemic dilated cardiomyopathy, Turitto et al(71) concluded that there is a strong correlation between an abnormal SAECG and induced ventricular tachyarrhythmia; however, both findings were uncommon in this subset of patients with only 15 percent exhibiting an abnormal SAECG.
Vester et al(72) in a study of 157 patients with dilated cardiomyopathy found a positive predictive value of only 24 percent and a negative predictive value of 93 percent.
In a group of at-risk patients with idiopathic dilated cardiomyopathies (n = 84), Yi et al(28) found that spectral turbulence analysis of SAECG (a newer frequency domain technique that detects frequent and abrupt changes in the QRS, permitting the assessment of the entire QRS complex) was no better than chance in predicting death (or aborted death) or progressive heart failure (54 percent, 50 percent, respectively). Of those patients with idiopathic dilated cardiomyopathies whose spectral turbulence analysis was considered abnormal (31/84, 37 percent), 18/31 (57 percent) experienced an adverse outcome during followup (mean 24 ± 18 months, range 1-59 months). Of the patients who did not remain clinically stable during followup (28/84, 34 percent), most had progressive heart failure (24/28, 85 percent), and the remainder died (4/28, 15 percent). Of those who survived, the spectral turbulence analysis was abnormal in 62 percent (15/24) of patients with idiopathic dilated cardiomyopathy. Of those who died, 75 percent were abnormal (3/4). Multivariate analysis determined that peak oxygen consumption (<19 mL/kg/min) presented the largest relative risk (9.55, 95 percent confidence interval [CI] 2.1-43.9) for the development of progressive heart failure in idiopathic dilated cardiomyopathic patients. Multivariate analysis also demonstrated that peak oxygen consumption, left ventricular end-diastolic dimension, and left ventricular ejection fractions were more likely to predict progressive heart failure than spectral turbulence analysis (relative risks were 6.43 [95 percent CI 1.28-32.4], 4.08 [95 percent CI 1.46-11.37], 3.35 [95 percent CI 1.01-1.28], and 3.2 [95 percent CI 1.04-9.08], respectively).
Ventricular tachyarrhythmia is known to be associated with an increased risk of cardiac death in the postinfarction period, with mortality rates approaching 60 percent. Proponents of SAECG have attempted to show that it is useful for risk stratification during recovery from MI. Many case series have reported SAECG, exercise testing, Holter monitoring, MUGA, programmed electrical stimulation, and coronary arteriography as useful for this purpose. In addition, supporting literature suggests radionuclide ventriculography to be a powerful predictor of survival. Other literature credits programmed electrical stimulation and Holter monitoring with being the best predictors of ventricular arrhythmic risk.
Results have been consistent in showing high negative predictive values (low false-negative rates) for adverse outcomes in post-MI patients with a negative SAECG. As mentioned previously, referral bias, case definitions, lack of randomization, and lack of masking of the investigators are probable methodologic flaws that may obscure generalizability of these results to different risk populations. Although other uses for SAECG have been proposed and carried out (such as in antiarrhythmic drug trials), the literature on post-MI patients is the most extensive to date.
The data supporting the use of SAECG in cardiomyopathic conditions were generated from small populations with ischemic and nonischemic conditions. Although a positive SAECG was not demonstrated to be an independent risk factor, two studies(28, 58) reported more morbidity (progressive heart failure, ischemia) in positive SAECG patients during followup. However, these studies were plagued by very high false-positive rates, which greatly diminished their value.
The data indicate that the SAECG in patients with cardiomyopathy has a weak positive predictive value and a strong negative predictive value. Conclusions about the value of SAECG in patients with cardiomyopathy, however, have been conflicting and may be explained in part by the variables in patient characteristics and the etiology of the myopathy.(13, 68, 70) Although SAECG has some prognostic value in patients with cardiomyopathy, a positive SAECG is a poor predictor of risk of sudden death in advanced cases. A 1995 review of the ability of SAECG to identify cardiomyopathy patients at high risk for sudden death concluded that, although SAECG is often abnormal in hypertrophic cardiomyopathy, it is not useful for risk stratification.(66)
Despite differences in SAECG component design and averaging techniques, several authors have demonstrated an increased risk of sudden death or ventricular tachyarrhythmia in SAECG-positive patients over patients who are SAECG-negative. The low sensitivity and poor predictive value of a positive SAECG present problems for clinicians in determining the utility of SAECG in both post-MI patients and those with cardiomyopathy.(50) Until the false-positive rate for SAECG can be reduced, independent confirmation of myocardial risk by other invasive or noninvasive methods appears to be prudent. SAECG alone appears unlikely to help in stratifying patients for further diagnostic or treatment interventions, but it does appear to serve for such discrimination after acute MI when combined with left ventricular ejection fraction.
Multivariate testing appears to be superior to univariate testing, and results of SAECG post-MI appear to differ depending on when it is performed. However, the optimal time to perform SAECG post-MI has not been determined.(1, 34, 44, 49, 53, 56, 58)
Based on the clinical data submitted to the Food and Drug Administration (FDA), through premarket notifications, the evidence to justify the claim that the performance of SAECG affects health outcomes is insufficient. However, based on expert consensus (e.g., the American College of Cardiology expert consensus document published in J Am Coll Cardiol 1996;27:238-249), the FDA believes that, for patients recovering from MI, SAECG has proven negative predictive accuracy when used to assess risk for development of life-threatening ventricular arrhythmias. The positive predictive accuracy of SAECG alone is not yet established.
The National Institutes of Health believe that SAECG is a safe and effective technology that can be clinically useful when combined with other standard tests to stratify risk of ventricular tachycardia in post-MI patients. The negative value has clinical significance because it can obviate the need for additional tests. Further study is required to confirm the use of SAECG in the surgical treatment of ventricular arrhythmias, detection of cardiac transplant rejection, early recognition of reperfusion after acute MI, and monitoring of antiarrhythmic drug treatment.
A wide variety of signal-averaging methods continue to be used. However, time domain is the primary tool for SAECG analysis. Although definitions of late potentials are still not standardized,(73) their identification has improved, and their prognostic value for clinically significant arrhythmias has been demonstrated after MI. Prognostication is enhanced when late potential analysis is combined with other tests such as left ventricular ejection fraction, Holter monitoring, or programmed electrical stimulation.
Several consistent results have been observed from the SAECG clinical studies: 1) a very high negative predictive value (76-100 percent); 2) variable sensitivity (35-83 percent) and specificity (47-91 percent); and 3) poor positive predictive value (8-48 percent) in at-risk populations. Many studies have demonstrated higher adverse effects in patients with positive VLPs or positive SAECG after MI than in those with negative VLPs. The prevalence of VLP was highest at 6 weeks to 3 months post-MI.
Given the inherent differences and lack of standardization in operating principles, institutional variances in methodology, and patient selection criteria, the routine use of SAECG as a univariate indicator of risk remains to be proven. In addition, the reproducibility of SAECG has been problematic.(74) Efforts that have been underway for the past 10 years in attempting to establish and sustain a unique role for SAECG technique have not been successful, despite the high negative predictive values, which can reduce the need for additional testing in patients with other single risk factors.
The preponderance of evidence suggests that other diagnostic measurements, such as left ventricular ejection fraction and programmed electrical stimulation, are useful in identifying patients at risk for ventricular tachyarrhythmia and sudden death. The acute MI and the cardiomyopathy literature have repeatedly demonstrated that the functional state of the left ventricle, whether compromised by ischemic or nonischemic processes, reliably predicts future risk (sudden death, ventricular tachyarrhythmia).(28) However, none of these modalities is sufficiently accurate as a univariate predictor or consistently better than SAECG. These modalities are perhaps most appropriately used in combination.(27, 34, 49, 53, 58)
A negative SAECG in the immediate post-MI period is associated with a lower morbidity and mortality in the first year post-MI. However, any conclusions about the prognostic value or utility of SAECG alone should await the results of ongoing prospective clinical trials on SAECG.
Patients with a positive SAECG have other risk factors, such as coronary artery disease, ventricular tachyarrhythmia, reduced left ventricular function, ventricular tachyarrhythmia on Holter monitoring, and other inducible arrhythmias. The high false-positive SAECG rate may obscure its utility when used alone. A positive SAECG, when combined with other standard tests, has been demonstrated to have clinical utility in post-MI patients. Other patient populations have not been conclusively shown to benefit from its use.
