Risk Stratification After Myocardial Infarction Using Signal-Averaged Electrocardiographic Criteria Adjusted for Sex, Age, and Myocardial Infarction Location
Background The objectives were to investigate the factors influencing signal-averaged ECGs (SAECGs) recorded in patients after myocardial infarction (MI) and to develop criteria for predicting arrhythmic events (AEs) that account for these factors.
Methods and Results SAECGs were recorded 5 to 15 days after MI in 2461 patients without bundle-branch block. The duration (QRSd), terminal potential (VRMS), and terminal duration (LAS) of the filtered QRS were measured. During follow-up (17±8 months), AEs (arrhythmic death; ventricular tachycardia, VT; ventricular fibrillation, VF) occurred in 80 patients (3.3%). Receiver operating characteristic curves showed that QRSd discriminated patients with all types of AEs, but VRMS and LAS discriminated only VT patients; QRSd minus LAS also discriminated AE patients. Sex, age, and MI location significantly affected the SAECG; survivors without VT or VF were divided into subgroups (2 sex×4 age×2 MI), and QRSd values exceeding the 70th percentile in each subgroup predicted AEs with a sensitivity of 65.4%. An unadjusted QRSd criterion showed the same overall sensitivity and specificity but with less uniform values for each subgroup. A Cox model was constructed by use of multiple prognostic indicators, and in rank order, QRSd, previous MI, and Killip class were predictive of AEs.
Conclusions SAECG adjustments for sex, age, and MI location did not improve sensitivity and specificity but produced a more uniform predictive performance. The proposed criteria are based only on QRSd, because late potentials (VRMS and LAS) did not discriminate patients with sudden death. Duration of high-level activity during QRS (QRSd−LAS) can predict AEs, suggesting that the arrhythmogenic substrate involves a large mass of myocardium.
The SAECG has been identified in numerous prospective studies1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 as one of the most powerful diagnostic tools for the assessment of patients at increased risk of ventricular tachyarrhythmias after MI.16 Sex and age have a significant influence on the standard ECG of normal subjects and are taken into account for defining the limits of normality.17 However, sex and age are not included in SAECG criteria.18 The QRSd measured on the SAECG is longer in men than in women in both healthy subjects19 20 21 and MI patients,22 and significant correlations have been found between age and the SAECG in MI patients.23 Also, in MI patients, it is not possible to accurately define late potentials with respect to the SAECG of normal subjects, because the activation sequence is modified by the infarct. Activation around the damaged wall occurs later in inferior MI than in anterior MI, which results in a longer QRSd and a higher incidence of late potentials in patients with inferior MI.3 24 25 SAECG criteria should thus be defined with respect to MI location in arrhythmia-free MI survivors.
The CAMI study was undertaken in 1990 to investigate the impact of new interventions on in- and out-of-hospital events, patient profile, and risk stratification methods.26 The number of patients enrolled in this study (n=2461) exceeds those of previous prospective SAECG studies1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 (n=102 to 1158), thus enabling multivariate analysis to be performed so that all factors19 20 21 22 23 24 25 that could affect the SAECG can be investigated. This large number of patients also allows the division of the population into subgroups for the definition of SAECG criteria adjusted for factors such as sex, age, and MI location.
All consecutive patients in eight Canadian hospitals who had sustained acute MI and met the selection criteria were considered for entry into the prospective study. Recruitment took place between June 1, 1990, and June 30, 1992, in six centers and between June 1, 1990, and December 31, 1992, in three centers. Patients were identified by daily review of all hospital admissions to the coronary care or intensive care units and to the cardiology services. In addition, cardiologists were questioned for identification of other patients who had had an acute MI but were hospitalized elsewhere in the hospital. The study design was approved by the institutional review boards of each center.
The diagnosis of acute MI required at least two of the following criteria: (1) characteristic chest pain lasting ≥20 minutes, (2) new Q wave with a duration of >40 ms evolving in the postadmission series of ECGs in at least two contiguous leads, and (3) creatine kinase ≥1.5 times the upper limit of normal and/or creatine kinase MB fraction >5% when simultaneous reference creatine kinase exceeded the upper limit of normal.
Of a total of 4133 patients with acute MI who were documented,26 379 patients died in the hospital before the SAECG could be recorded; of the remainder, the following groups of patients satisfied the exclusion criteria: 232 patients who were >75 years old at enrollment time, 37 patients who had malignant neoplasm, 95 patients who were in critical or unstable condition during the SAECG recording time window, 286 patients who refused to participate or whose physician refused, 131 patients because of geographic remoteness that would impair the follow-up, and 205 patients with bundle-branch block or permanent pacemakers. In addition, the following patients were excluded even if they were eligible: 302 patients because of technical reasons (equipment failure, holidays) or patient communication problems and 5 patients with an incomplete file or lost to follow-up. The remaining 2461 patients were included in the SAECG study.
All included patients had demographic variables and cardiac history assessed. In addition, the clinical course of their MI (heart failure, ventricular arrhythmias, recurrent MI, etc) and the therapeutic interventions (thrombolytic treatment, coronary angioplasty, coronary bypass, drugs, etc) were documented. The diagnosis of recurrent MI, if nonfatal, required the same criteria as the entry criteria of the study itself and if fatal, had to meet the criteria of the mortality committee (see below). The clinical management of patients was at the discretion of individual physicians who were not aware of the SAECG results.
The SAECG was recorded 5 to 15 days after MI, because SAECGs recorded during this period have been identified as having the most significant relation to arrhythmic events occurring during the first year.5 Three bipolar orthogonal leads were used. After skin preparation, electrodes were applied over the left and right midaxillary lines at the V5 level (+X and −X), along the left midclavicular line over the infraclavicular area and the left iliac fossa (+Y and −Y), and over the cardiac apex and the corresponding position on the back (+Z and −Z). Recordings were performed in the supine position with Predictor I systems (Corazonix). The sampling rate was 2000 s−1 and the bandwidth was 0.05 to 300 Hz. The sampling window had a width of 300 ms. The beats were aligned by use of an updated template to prevent the degradation of the averaging process.27 Sufficient beats (≈400) were averaged to achieve a noise level of 0.3 μV in all SAECG recordings.28 Noise was calculated as the mean of the root-mean-square noise of the averaged X, Y, and Z leads during the 50 ms at the end of the acquisition window.
The processing of all SAECG recordings was performed at the coordinating center. Bidirectional high-pass filtering was repeated for three different high-pass cutoff frequencies: 25, 40, and 80 Hz. In the rest of this article, SAECG results are presented for 40 Hz, which showed the highest sensitivity for predicting arrhythmic events (see “Results”). QRS onset and offset were determined automatically (Predictor I, version 5.0 software) and visually validated. In rare instances (1% to 2%) when the end points could not be automatically detected or were not correctly detected, eg, because the P wave was smeared into the QRS complex, the end points were set manually. Three variables were computed (see Fig 1⇓): QRSd, LAS, and VRMS. In addition, QRSd−LAS was also computed.
MI location, as assessed by 12-lead ECG, was classified into three categories: (1) non–Q-wave, (2) inferoposterior, and (3) anterolateral. Of the 2461 patients included, 2064 (80%) had a technically adequate measure of the LVEF. Ambulatory ECGs were recorded and judged satisfactory (>18 hours) in 1803 patients. After automated full disclosure analysis followed by physician confirmation, the average frequency of PVCs (PVCs per hour) and the occurrence of nonsustained ventricular tachycardia were recorded.
Included patients had follow-up visits after 6 and 12 months after MI to record new SAECGs and to complete a questionnaire documenting the evolution and treatment of their disease. At 24 months and also if a patient could not return to the participating center for a new SAECG measurement, the questionnaire was completed by phone. This questionnaire documented cardiac interventions and complications (eg, angina, MI, syncope, arrhythmias) that occurred during the follow-up period as well as the current status of the patient (eg, Goldman scale, smoking, exercise). Excluded patients were phoned after 1 and 2 years to assess their vital status. All cardiovascular events were verified by chart review or by patient records obtained from other hospitals and/or physicians.
The details concerning the deaths of all included and excluded patients were ascertained at the participating centers. A central mortality committee, blinded with regard to SAECG results, then reviewed the information and classified deaths according to predefined criteria. Deaths were first classified according to whether they were cardiovascular or noncardiovascular. Cardiovascular deaths were classified into (1) recurrent MI, subclassified as definite according to the same criteria as used for the qualifying MI or proven at autopsy, or probable if the patient died within the same hospital admission or within 7 days of recurrent MI; (2) pump failure, defined as death in a patient with documented NYHA class IV failure or shock with at least two of the following three criteria: evidence of poor perfusion, blood pressure <90 mm Hg, or refractory congestive symptoms; (3) arrhythmic death, subclassified as definite or presumed according to the CAST criteria29 ; (4) procedural, defined as an unexpected death directly resulting from a complication related to a cardiovascular procedure; and (5) other vascular deaths (eg, stroke, aneurysms, or peripheral or renovascular disease). When a death resulted from an acute MI, it was classified as such whether the final event was low output or arrhythmias.
Study End Points
Three specific arrhythmic events were defined a priori as study end points: (1) arrhythmic death, as defined above29 ; (2) sustained VT, defined as a documented tachycardia of ventricular origin with a rate ≥120 min−1 and lasting ≥30 seconds; and (3) documented VF requiring defibrillation. When a patient experienced more than one event, the first was taken as the study end point.
Continuous variables are expressed as mean±SD. Univariate comparisons between groups were conducted with Student’s t test for continuous variables, the Mann-Whitney U test for continuous variables when the normality assumption was questionable, and χ2 analysis for categorical variables. The type I error was fixed at P=.05.
An ANCOVA was used to assess the effects of MI location, sex, and thrombolytic treatment factors; of the interactions between these factors; and of covariates such as age and LVEF on the QRSd, the LAS, and the logarithm of the VRMS. A logarithmic transformation was applied to the VRMS to normalize its distribution. A contrast analysis was used to detect differences between MI locations, and the Bonferroni method was applied to adjust for multiple comparisons.
ROC curves were computed to assess the sensitivity and specificity of the QRSd, VRMS logarithm, and LAS variables for predicting arrhythmic events during follow-up. The Hanley-McNeil test was used to test and to compare the discrimination power of these variables.30
The Cox proportional-hazards regression model was applied in a stepwise manner (both forward and backward) to evaluate the power of SAECG variables, LVEF, Holter variables, and other clinical features that are known predictors of survival after MI.26 The relation between arrhythmic end points and the SAECG test was assessed with the Kaplan-Meier product-limit estimate of the survival function. The statistical tests were performed with the Systat software (version 5.01 DOS), with the exception of the Hanley-McNeil test, which was performed with a custom program.
Follow-up of Included and Excluded Patients
The mortality of included and excluded patients is shown in Table 1⇓. The overall mortality for patients who were ineligible because of cancer or critical condition was significantly higher than for the included patients (28% versus 8.8%). Mortality in patients ineligible because of bundle-branch block or pacemaker was also higher. There was no significant difference in mortality between included patients and those excluded because of patient or physician refusal and for geographical or technical reasons. Among included patients, the mean follow-up duration in the survivors was 17±8 months (range, 12 to 24 months). After the SAECG recording, 217 patients (8.8%) died, 143 (5.8%) of cardiovascular causes. Eighty patients (3.3%) had a primary end point during follow-up: arrhythmic death in 51 patients (2.1%), sustained VT in 21 (0.9%), and resuscitated VF in 8 (0.3%).
The clinical characteristics of patients with and without an arrhythmic event during follow-up are summarized in Table 2⇓. The two groups were similar with respect to age, sex, MI location, and the percentages of patients who underwent coronary artery bypass graft surgery or angioplasty during their hospitalization. Patients with previous MI or angina, those with a lower LVEF or heart failure (Killip class III and IV), and those who had not received thrombolytic treatment were more likely to have subsequent arrhythmic events.
Factors Affecting the SAECG
An ANCOVA was performed to assess the effects of MI location, sex, thrombolytic treatment, age, and LVEF on the SAECG variables. Table 3⇓ summarizes the ANCOVA results for the 2064 patients who had LVEF data (note, there was no significant difference in the incidence of arrhythmic events in patients with and without LVEF data: 3.8% and 3.1%, respectively).
For the QRSd, a significant difference in the adjusted group means was found for each of the three independent variables: MI location, sex, and thrombolytic treatment (P<.001). A significant correlation was also found between QRSd and the two covariates (age and LVEF). The F ratios indicate that sex and LVEF had the strongest influence on QRSd, whereas thrombolytic treatment had the weakest influence. Contrast analysis shows that there was no significant difference in the QRSd between non–Q-wave and inferoposterior MI locations, whereas there was a very significant difference between either of these MI locations and anterolateral MI.
Some antiarrhythmic drugs, specifically sodium channel blockers, are known to affect the SAECG.31 In this study, only 78 patients (3.1%) received an antiarrhythmic drug within five half-lives before the SAECG recording. The SAECGs of patients receiving amiodarone, procainamide, and quinidine showed longer QRSd and LAS than those of patients who had not received any drug, whereas no significant differences were found for patients receiving sotalol and verapamil. Thus, antiarrhythmic drugs were infrequently prescribed and could have modified the SAECG in only a few patients in this study.
Risk Stratification With Unadjusted SAECG Criteria
The presence of late potentials is defined by prolonged QRSd and LAS and reduced VRMSs. ROC curves representing the sensitivity versus specificity obtained when different thresholds were applied to the SAECG variables for the prediction of all arrhythmic events are shown in Fig 2⇓. In addition to characterizing the interplay between sensitivity and specificity, ROC curves can also be used to compare different tests by measuring the area under the curve, which reflects the discrimination power of the variable. All three SAECG variables as well as a variable representing the duration of high-level activity during the QRS complex obtained by subtracting the LAS from the QRSd (QRSd−LAS) showed a significant discrimination power for predicting all arrhythmic events. However, the discrimination powers of QRSd and QRSd−LAS were significantly higher than those of LAS and VRMS, and there was no significant difference between LAS and VRMS and between QRSd and QRSd−LAS.
To further investigate the poorer performance of LAS and VRMS for predicting all arrhythmic events, ROC curves were plotted for the prediction of specific arrhythmic events. For predicting sustained VT (Fig 2⇑, right), all four SAECG variables showed a significant discrimination power, but the discrimination power of QRSd was the highest. For the prediction of arrhythmic death (Fig 3⇓, left), only the QRSd and the QRSd−LAS showed a significant discrimination power, with no significant difference between QRSd and QRSd−LAS. For the prediction of VF (Fig 3⇓, right), only the QRSd showed a significant discrimination power.
QRSd Percentile Distributions of Arrhythmia-Free Survivors
SAECG criteria adjusted for sex, age, and MI location were developed for the QRSd variable alone because it could predict all types of arrhythmic events, whereas the LAS and VRMS could not predict arrhythmic death or VF. To develop a reference interval, percentile values describing the distribution of the QRSd were computed for the 2174 arrhythmia-free survivors who did not receive any antiarrhythmic drugs before the SAECG recording, taking into account the three factors that significantly affected the QRSd: sex, age, and MI location. The non–Q-wave and the inferoposterior MI locations were combined because of the lack of difference between those two MI locations. To take age into account, patients were divided into four age groups: <50, 50 to 59, 60 to 69, and 70 to 75 years old.
Figs 4⇓ and 5⇓ present the QRSd percentile curves with respect to age group and MI location for male and female arrhythmia-free survivors, respectively. The median (50%) QRSd was longer in men than in women, in non–Q-wave and inferoposterior MI than in anterolateral MI, and in older patients. A similar relation was observed for the median LAS (not shown), whereas the median VRMS (not shown) was smaller in men than in women, in inferoposterior MI than in non–Q-wave and anterolateral MI, and in older patients. Since the upper percentile values can be used to define an abnormally prolonged QRSd, these values are also given in Table 4⇓.
Risk Stratification With Group-Specific SAECG Criteria
The data presented in Table 4⇑ were used to identify patients exceeding the 70th percentile for a given subgroup and to predict arrhythmic events occurring during the follow-up in the entire population (n=2461). The 70% specificity value was selected on the ROC curve because it represents a reasonable compromise between sensitivity and specificity. Note that the combination of the ROC curve (Fig 2⇑, left) and the QRSd distributions (Table 4⇑) gives the information needed to assess the sensitivity and specificity for other cutpoints.
Table 5⇓ presents the sensitivities and specificities observed for predicting arrhythmic events for each subgroup according to sex and MI location. In addition, the sensitivities and specificities were also computed for an unadjusted criterion that corresponds to the 70th percentile of the QRSd for the group of arrhythmia-free and drug-free survivors (QRSd >109 ms). The group-specific QRSd criteria predicted arrhythmic events with an overall sensitivity of 65.0% and an overall specificity of 68.4%. These values were almost identical to the values obtained with the simpler unadjusted criterion (65.0% and 69.2%, respectively). However, the sensitivities and specificities observed for the different subgroups were less uniform for the unadjusted criterion. For example, the unadjusted criterion showed a sensitivity ranging between 36.4% for women with non–Q-wave or inferoposterior MI and 77.8% for men with non–Q-wave or inferoposterior MI; the corresponding range was 45.5% to 71.1% for the group-specific criteria. The specificities ranged between 63.5% and 85.6% for the unadjusted criterion and remained at 68% to 69% for the group-specific criteria. The specificities for the group-specific criteria were very close but not equal to 70%, because the percentile distributions were computed for arrhythmia-free and drug-free survivors, whereas the SAECG specificity was evaluated for all patients without arrhythmic events.
To test the uniformity of the sensitivity of the group-specific criteria, the QRSd value of each patient with an arrhythmic event was first ranked with respect to the subgroup of arrhythmia-free and drug-free survivors to which he or she belonged. Each QRSd value was thus transformed into a percentage value. A nonparametric Mann-Whitney U test then showed that there was no significant difference between the percentage values of men versus women (74±27% versus 64±30%), and a nonparametric Kruskal-Wallis ANOVA showed that there was no significant difference between the percentage values for the non–Q-wave, inferoposterior, and anterolateral MI locations (72±26%, 76±32%, and 70±29%, respectively) and for the different age groups (76±37%, 73±35%, 75±25%, and 62±29% for groups of increasing age, respectively). In addition, group-specific QRSd criteria were computed for filter frequencies of 25 and 80 Hz and showed sensitivities that were lower (60.0% and 63.1%, respectively) than the sensitivity obtained at 40 Hz, which justifies the choice of the 40-Hz filter frequency.
Relationship Between the SAECG and Other Variables for Risk Stratification
The Cox proportional-hazards model was applied in a stepwise manner to evaluate the power for predicting arrhythmic events of the following variables: continuous SAECG variables (QRSd, LAS, log VRMS), LVEF (continuous), Holter variables (>10 PVCs per hour dichotomized; occurrence of nonsustained VT), and clinical features that have been associated with out-of-hospital survival during the first year after MI in this population26 (diabetes mellitus, Killip class, previous coronary artery bypass, previous MI, in-hospital coronary angioplasty, thrombolytic treatment, in-hospital coronary artery bypass). For both forward and backward stepwise procedures, three variables were shown to have a significant and independent contribution for the prediction of arrhythmic events. In order of selection, these variables were (1) QRSd, (2) history of previous MI, and (3) Killip class. After the patients for whom the LVEF and/or the ambulatory ECG was not available were included, the same three variables were retained by the stepwise procedure. The LVEF, which was a univariate predictor of arrhythmic events, did not reach statistical significance after the inclusion of the Killip class.
Table 6⇓ summarizes the performance of the three variables, alone or in combination, for the prediction of arrhythmic events. For these variables alone, a prolonged QRSd had the highest sensitivity (65.0%) and the lowest specificity (68.4%), whereas a Killip class 3 to 4 had the highest specificity (93.5%) and the lowest sensitivity (17.5%). For all three variables, the positive-predictive accuracies are low (6.5% to 8.3%) and the negative-predictive accuracies are very high (97.1% to 98.3%) because of the low event rate. When considered individually, a prolonged QRSd had the highest risk ratio (3.8) of the three variables. For the different combinations of variables, a prolonged QRSd with a previous MI had the highest risk ratio (4.8) but a positive-predictive accuracy that remains low (10.3%).
Finally, the time occurrence of arrhythmic events after MI was investigated by plotting Kaplan-Meier survival curves (Fig 6⇓) showing the percentage of patients that remained arrhythmia-free as a function of time after MI for patients with and without a prolonged QRSd by group-specific criteria (>70%). The difference between the two survival curves is highly significant (P<.0001).
Factors that can affect the SAECG were first investigated in a large population of MI patients by ANCOVA. This analysis revealed some significant relationships between the SAECG and sex, age, MI location, and LVEF. These relatively weak relationships were not detected in previous prospective studies with smaller patient populations for which no significant differences in sex distribution,4 5 6 average age,4 5 6 MI location distribution,5 6 7 10 and LVEF5 6 7 10 were found between MI patients with a normal SAECG and those with an abnormal SAECG. The longer QRSd observed in men can be attributed to a larger mass of myocardium, which is in turn related to a larger body size, as suggested by the cancellation of these differences after normalization for height.20 A significant increase of QRSd with age was also reported in MI patients by Malik et al23 and could be related to degenerative processes affecting conduction.32 The shorter QRSd observed in patients with anterolateral MI3 4 24 25 could be explained by the activation around the damaged wall, which occurs earlier in anterior MI.
Prediction of Arrhythmic Events With Group-Specific SAECG Criteria
This prospective study shows that the SAECG can predict arrhythmic events. Even if factors such as sex, age, and MI location significantly affect the SAECG, adjustments of SAECG criteria for these factors did not improve the overall sensitivity and specificity. However, these adjustments equalized the predictive performance for the different subgroups, which was not uniform with unadjusted criteria.3 24 25 These adjustments are important for the clinical application of the SAECG.
SAECG Differences for VT, VF, and Arrhythmic Death
Because of the large patient population among which sufficient numbers of VTs, VFs, and arrhythmic deaths occurred, different features of the SAECG could be associated specifically with the prediction of VT and not with that of VF or arrhythmic death. This was not possible in previous SAECG prospective studies, which combined all types of ventricular tachyarrhythmias because of their lower numbers of events.1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Thus, the group-specific SAECG criteria that are proposed are based only on the QRSd because of the inferior predictive performance of the VRMS and LAS variables. The predictive performances of VRMS and LAS were acceptable only for VT patients; they were not significant for VF patients and were very poor for patients with arrhythmic death. The VRMS and LAS variables, along with the QRSd, were initially proposed by Simson33 to specifically identify VT patients in a retrospective study of MI patients, and differences in the SAECG measured in MI patients with either VT or VF were reported later.34 35 36 37 38 Thus, in resuscitated survivors of spontaneous VF, the VRMS was found to be larger34 36 and the QRSd shorter35 than in patients with recurrent sustained VT, whereas the VRMS was similar to that of normal subjects.34 In MI patients who underwent an electrophysiological study, Martinez-Rubio et al37 reported that VRMSs were larger and the QRSd shorter in patients with inducible VF than in patients with inducible VT. In their prospective studies, Gomes et al,3 Steinberg et al,8 and Pedretti et al14 found that only the QRSd was significantly associated with arrhythmic events, and Odemuyiwa et al38 reported some differences between the predictive characteristics of the different SAECG variables for arrhythmic death and VT.
Whereas the electrophysiological basis of late potentials in VT patients is a slow, delayed, and fragmented activation in the vicinity of the MI scar39 40 that may constitute a substrate for reentry circuits,41 the electrophysiological basis of the prolonged QRSd and its possible arrhythmogenic mechanism in patients with VF and arrhythmic death remains to be determined. The fact that the duration of high-level activity during the QRS complex (QRSd−LAS, excluding late potentials) had the same discrimination power for all arrhythmic events as the QRSd (including late potentials) suggests the involvement of a large mass of ventricular myocardium. Because of its high amplitude, this QRSd prolongation can be detected on the 12-lead ECG, which can give prognostic information similar to that of the SAECG.42 The QRSd prolongation could be explained by factors such as intraventricular conduction defects; ventricular dilatation, which elongates the ventricular conduction pathways without altering conduction velocity43 44 ; and ventricular remodeling, which increases tissue mass45 46 and could slow conduction velocity by altering cellular coupling. The hypothesis of a prolonged ventricular activation in dilated or remodeled ventricles is supported by the uniformity of the performance of the group-specific QRSd criteria: a shorter QRSd cutoff is necessary in female patients because they have smaller hearts, whereas a longer QRSd cutoff is necessary in older patients or in patients with inferoposterior MI because conduction velocity is already slower or the activation sequence is prolonged by the MI location in these groups of patients. Zaman et al47 also reported a correlation between the QRSd measured after MI and the changes in the end-diastolic volume index occurring weeks later.
Dilated48 or remodeled49 ventricles could thus constitute a specific VF substrate. For example, a significant proportion of the deaths occurring after hospital discharge and classified as arrhythmic deaths may not be caused by reentrant VT occurring in a chronic MI setting but rather by ventricular tachyarrhythmias produced by abnormal automatism resulting from an acute recurrent MI. In our patient population,50 autopsy was performed in 22 of the arrhythmic deaths and revealed the presence of a fresh, macroscopic thrombus in 5 cases (23%). This may explain in part the lack of sensitivity of late potentials for predicting arrhythmic death, because they reflect the presence of an abnormal peri-infarction activation that is potentially arrhythmogenic and not that of an arrhythmogenic focus induced by acute ischemia. On the other hand, a prolonged QRS could still play a role in the prediction of arrhythmic death associated with acute ischemia, because it may reflect an increased susceptibility to VF, whatever the mechanism of the initiating tachyarrhythmia.
Other Prognostic Indicators
The Cox proportional-hazards regression model was applied to evaluate the predictive value of SAECG variables and a dozen other clinical features that are known predictors of survival after MI. A prolonged QRSd was ranked highest among the independent predictors of arrhythmic events. Previous prospective studies that have compared the SAECG with other prognostic indicators have also determined that the abnormal SAECG,3 4 5 8 9 10 13 15 51 and more specifically the QRSd,3 8 9 14 15 is one of the best independent predictors of arrhythmic events. Two other independent predictors were identified: a history of previous MI and a high Killip class. Both of these support the hypothesis of a VF substrate formed by enlarged or remodeled ventricles: the history of previous MI could reflect the negative impact of a more extended and structurally complex remodeled myocardium, whereas the high Killip class reflects ventricular dysfunction and heart failure caused by remodeling that accounts for abnormal myocardial stiffness and pathological hypertrophy.52
A prolonged QRSd, a previous MI, and a high Killip class thus identify a group of patients with a higher risk of arrhythmic events. The risk ratio rises from 3.8 for a prolonged QRSd alone to 4.8 when the abnormal QRSd is combined with a history of previous MI; however, the positive-predictive accuracy, which rises from 6.5% to 10.6% when these variables are combined, remains low because of the low incidence of events, whereas the negative-predictive accuracy remains very high (>97%).
Role of the SAECG in Patient Management
The value of the SAECG for predicting arrhythmic events after MI clearly outranks that of more expensive tests such as LVEF measurements and ambulatory electrocardiography, but how could, or how should, the SAECG affect patient management? The answer depends on the patient population. In the CAMI study, because of the low incidence of events, the SAECG performed very well for identifying low-risk patients but poorly for identifying high-risk patients. The incidence of events being a critical factor in the predictive performance of the SAECG, how could these results be applied to another patient population? Fig 7⇓ illustrates the incidence of arrhythmic events occurring after discharge in the CAMI study and in 15 other prospective SAECG studies carried out over the past 15 years.1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 Even if these studies have different follow-up durations and exclusion criteria, a decreasing trend in the incidence of arrhythmic events can be observed. The positive-predictive accuracy of an abnormal SAECG ranged between 23% and 28% in the earlier studies1 2 3 4 5 and decreased to 11% to 17% in the latest studies.13 14 15 A similar decreasing trend in the post–hospital discharge mortality of MI patients occurred during the last decades.26 The widespread use of new therapeutic and interventional advances such as thrombolytic treatment, β-blocker agents, antiplatelet agents, and revascularization procedures or changes in lifestyle and in the management of risk factors may have improved postinfarction survival. This reduction of the risk challenges the clinical utility of all risk-stratification procedures because it lowers their positive-predictive accuracy. In the context of this decreasing trend and considering the proarrhythmic effects of antiarrhythmic drugs as highlighted in the CAST study,29 it is clinically relevant to accurately identify low-risk patients who would only be adversely affected by the proarrhythmic action of these drugs. Thus, because of its high negative-prediction accuracy, the SAECG could play an important role for patient selection in future interventional studies on the efficacy of antiarrhythmic agents or of automatic implantable cardioverter-defibrillators. For patient management in the nineties, a positive SAECG test indicates an important increase of the relative risk of arrhythmic events, but the role of the SAECG as a screening test is otherwise severely limited by its poor positive-predictive accuracy.
The QRSd percentiles reported in this study are directly applicable to other signals recorded with the same leads and analyzed with the same algorithms. The accuracy and reproducibility of these QRSd percentiles should be high, considering that the accuracy of QRS end point detection by computer algorithms increases with the signal-to-noise ratio and that all the recordings had a very low noise level. Finally, since the choice of the QRSd cutpoints is derived from the same population on which the predictive performance is also assessed, the performance of the test is likely to degrade when it is applied to another patient sample.
Even if such factors as sex, age, and MI location significantly affect the SAECG, adjustments of SAECG criteria for these factors do not improve the overall sensitivity and specificity for predicting arrhythmic events after MI. However, these adjustments effectively equalize the SAECG predictive performance for the different subgroups, which is an important clinical advantage. In contrast with previous SAECG studies that used criteria based on QRSd, LAS, and VRMS, we are recommending group-specific criteria for predicting arrhythmic events that are based only on prolonged QRSds, because low VRMS and long LAS values, which are typical of late potentials, are associated only with VT and not with VF or arrhythmic death. The duration of high-level activity during the QRS complex (QRSd−LAS, which excludes late potentials) had the same discrimination power for arrhythmic events as the QRSd (which includes late potentials), suggesting the involvement of a large mass of ventricular myocardium in the arrhythmogenic substrate. When multiple prognostic indicators are considered, a prolonged QRSd, a previous MI, and a high Killip class identify independently a group of patients with a higher risk of arrhythmic events after MI. The investigation of the electrophysiological basis and the possible arrhythmogenic mechanism of the prolonged QRSd found in patients with VF and arrhythmic death constitutes a future direction for research in this area.
Selected Abbreviations and Acronyms
|CAMI||=||Canadian Assessment of Myocardial Infarction|
|LAS||=||duration of terminal portion of QRS with vector magnitude <40 μV|
|LVEF||=||left ventricular ejection fraction|
|PVC||=||ventricular premature contraction|
|QRSd||=||total duration of filtered QRS complex|
|QRSd−LAS||=||duration of high-level activity during QRS|
|ROC||=||receiver operating characteristic|
|VRMS||=||root-mean-square voltage of vector magnitude during last 40 ms of QRS complex|
This study was financially supported through a University-Industry grant from the Medical Research Council of Canada and Bristol-Myers Squibb of Canada. The present study is an ancillary trial of CARE. The authors are grateful to Ann Baker, Vidoslav Balnozan, Danielle Beaudoin, Bonnie Cochrane, Catherine Dunne, Colette Favreau, Sheila Griffiths, Jeannette Kenny, Eric Lavallée, Charlotte Lavoie, Etel Mikes, Doris Morissette, Francine Ouimet, Denise Pagé, Andrea Serpa, Brenda Smith, and Helen Tremayne for their devoted and expert assistance. In memoriam Lise Legendre.
Reprint requests to Pierre Savard, Centre de recherche, Hôpital du Sacré-Coeur, 5400 Gouin ouest, Montréal, Québec, Canada H4J 1C5.
- Received July 27, 1995.
- Revision received January 13, 1997.
- Accepted January 22, 1997.
- Copyright © 1997 by American Heart Association
Breithardt G, Schwartzmaier J, Borggreffe M, Haerten K, Seipel L. The prognostic significance of late ventricular potentials after myocardial infarction. Eur Heart J. 1983;4:487-495.
Denniss AR, Richards DA, Cody DV, Russel PA, Young AA, Cooper MJ, Ross DL, Uther JB. Prognostic significance of ventricular tachycardia and fibrillation induced at programmed stimulation and delayed potentials detected on the signal-averaged ECG of survivors of acute myocardial infarction. Circulation. 1986;74:731-740.
Gomes JA, Winters SL, Martinson M, Machac J, Stewart D, Targonski A. The prognostic significance of quantitative signal-averaged variables relative to clinical variables, site of myocardial infarction, ejection fraction and ventricular premature beats: a prospective study. J Am Coll Cardiol. 1988;13:377-384.
Richards DAB, Byth K, Ross DL, Uther JB. What is the best predictor of spontaneous ventricular tachycardia and sudden death after myocardial infarction? Circulation. 1991;83:756-763.
Farrel TG, Bashir Y, Cripps T, Malik M, Poloniecki J, Bennet ED, Ward DE, Camm AJ. Risk stratification for arrhythmic events in post-infarction patients based on heart rate variability, ambulatory electrocardiographic variables and the signal-averaged electrocardiogram. J Am Coll Cardiol. 1991;18:687-697.
McClements BM, Adgey AAJ. Value of signal-averaged electrocardiography, radionuclide ventriculography, Holter monitoring and clinical variables for prediction of arrhythmic events in survivors of acute myocardial infarction in the thrombolytic era. J Am Coll Cardiol. 1993;21:1419-1427.
El-Sherif N, Denes P, Katz R, Capone R, Mitchel LB, Carlson M, Reynolds-Haertle R, and the CAST Investigators. Definition of best prediction criteria of the time domain signal-averaged electrocardiogram for serious arrhythmic events in the postinfarction period. J Am Coll Cardiol. 1995;25:908-914.
Kjellgren O, Gomes JA. Current usefulness of the signal averaged electrocardiogram. Curr Probl Cardiol. 1993;18:363-418.
MacFarlane PW, Lawrie TDV, eds. Comprehensive Electrocardiology: Theory and Practice in Health and Disease. Volume 3, Appendix 1, Normal Limits. New York, NY: Pergamon Press; 1989:1441-1526.
Breithardt G, Cain ME, El-Sherif N, Flowers NC, Hombach V, Janse M, Simson MB, Steinbeck G. Standards for analysis of ventricular late potentials using high-resolution or signal-averaged electrocardiography: a statement by a task force committee of the European Society of Cardiology, the American Heart Association and the American College of Cardiology. J Am Coll Cardiol. 1991;17:999-1006.
Yang TF, Kennedy J, Macfarlane PW. Sex dependence of normal limits of the signal-averaged electrocardiogram. Br Heart J. 1993;69(suppl):65. Abstract.
de Chillou C, Sadoul N, Bruntz JF, Simon JP, Gazakur E, Aliot E. Importance of gender in the definition of an abnormal signal averaged electrocardiogram after myocardial infarction. Eur J Card Pacing Electrophysiol. 1994;4:207. Abstract.
Buckingham TA, Greenwalt T, Lingle A, Volgman AS, Koper P, Janosik D, Zbilut JP. In anterior myocardial infarction, frequency domain is better than time domain analysis of the signal-averaged ECG for identifying patients at risk for sustained ventricular tachycardia. Pacing Clin Electrophysiol. 1992;15:1681-1687.
Elko PP, Weaver WD, Kudenchuk P, Rowlandson I. The dilemma of sensitivity versus specificity in computer interpreted acute myocardial infarction. J Electrocardiol. 1992;24:2-7.
Rouleau JL, Talajic M, Sussex B, Warnica W, Davies R, Potvin L, Gardner M, Stewart D, Plante S, Dupuis R, Lauzon C, Ferguson J, Mikes E, Balnozan V, Savard P. Myocardial infarction patients in the nineties, their risk factors, stratification and survival in Canada. J Am Coll Cardiol. 1996;27:1119-1127.
Jones J, Srodulski ZM, Romisher S. The aging electrocardiogram. Am J Surg Med. 1990;8:240-245.
Simson MB. Use of signals in the terminal QRS complex to identify patients with ventricular tachycardia after myocardial infarction. Circulation. 1981;64:235-242.
Freedman RA, Gillis AM, Keren A, Soderholme-Difatte V, Mason JW. Signal averaged electrocardiographic late potentials in patients with ventricular fibrillation or ventricular tachycardia: correlation with clinical arrhythmias and electrophysiologic study. Am J Cardiol. 1985;55:1350-1353.
Denniss AR, Ross DL, Richards DA, Holley LK, Cooper MJ, Johnson DC, Uther JB. Differences between patients with ventricular tachycardia and ventricular fibrillation as assessed by signal-averaged electrocardiogram, radioventriculography and cardiac mapping. J Am Coll Cardiol. 1988;11:276-283.
Martinez-Rubio A, Shenasa M, Borggreffe M, Chen X, Benning F, Breithardt G. Electrophysiologic variables characterizing the induction of ventricular tachycardia versus ventricular fibrillation after myocardial infarction: relation between ventricular late potentials and coupling intervals for the induction of sustained ventricular tachyarrhythmias. J Am Coll Cardiol. 1993;21:1624-1631.
Odemuyiwa O, Malik M, Poloniecki J, Farrell T, Millane T, Kulakowski P, Staunton A, Matthies A, Camm AJ. Differences between predictive characteristics of signal averaged electrocardiographic variables for postinfarction sudden death and ventricular tachycardia. Am J Cardiol. 1992;69:1186-1192.
Gardner PI, Ursell PC, Fenoglio JJ, Wit AL. Electrophysiologic and anatomic basis for fractionated electrograms recorded from healed myocardial infarction. Circulation. 1985;72:596-611.
Lacroix D, Savard P, Shenasa M, Pagé P, Joly D, Derome D, Cardinal R, Nadeau R. Spatial domain analysis of late ventricular potentials: intraoperative and thoracic correlations. Circ Res. 1990;66:55-68.
Lacroix D, Savard P, Shenasa M, Kaltenbrunner W, Cardinal R, Pagé P, Nadeau R. Intracardiac and thoracic mapping of late ventricular potentials. In: Shenasa M, Borggrefe M, Breithardt G, eds. Cardiac Mapping. Armonck, NY: Futura Publishing Co; 1993:367-382.
Reiter MJ, Mann DE, Williams GR. Interaction of hypokalemia and ventricular dilatation in isolated rabbit hearts. Am J Physiol. 1993;265:H1544-H1550.
Reiter MJ, Zetelakis Z, Kirchhof JH, Boersma L, Allesie MA. Interaction of acute ventricular dilatation and d-sotalol during sustained reentrant ventricular tachycardia around a fixed obstacle. Circulation. 1994;89:423-431.
Pfeffer MA, Braunwald E. Ventricular remodeling after myocardial infarction. Circulation. 1990;81:1161-1172.
Zaman AG, Morris JL, Smyllie JH, Cowan JC. Late potentials and ventricular enlargement after myocardial infarction: a new role for high-resolution electrocardiography? Circulation. 1993;88:905-914.
Jalal S, Williams GR, Mann DE, Reiter MJ. Effect of acute ventricular dilatation on fibrillation thresholds in the isolated rabbit heart. Am J Physiol. 1992;263:H1306-H1310.
Hadjis T, Talajic M, Stewart D, Sampalis J, Sussex B, Warnica W, Davies R, Gardner M, Dupuis R. The role of thrombosis in sudden cardiac death: the SAGES trial experience. Circulation. 1994;90(suppl I):I-374. Abstract.
Mäkijärvi M, Fetsch T, Reinhardt L, Martinez-Rubio A, Borggrefe M, Breithardt G. A new diagnostic algorithm for noninvasive risk stratification after acute myocardial infarction. Circulation. 1994;90(suppl I):I-501. Abstract.
Brilla CG. The cardiac structure-function relationship and the renin-angiotensin-aldosterone system in hypertension and heart failure. Curr Opin Cardiol. 1994;9(suppl 1):S2-S10.