780
Combined Use of Time and Frequency Domain
Variables in Signal-Averaged ECG as a
Predictor of Inducible Sustained Monomorphic
Ventricular Tachycardia in
Myocardial Infarction
Akihiko Nogami, MD; Yoshito lesaka, MD; Junichi Akiyama, MD; Atsushi Takahashi, MD;
Junichi Nitta, MD; Yeong-hwa Chun, MD; Kazutaka Aonuma, MD; Michiaki Hiroe, MD;
Fumiaki Marumo, MD; and Masayasu Hiraoka, MD
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Background. Time and frequency domain analyses of signal-averaged ECG (SAECG) have several
individual limitations, and the results of the two methods sometimes vary considerably. The purpose of
this study was to determine whether the combined use of time and frequency domain variables facilitates
identification of patients who will have ventricular tachycardia (VT) induced during programmed
ventricular stimulation (PVS).
Methods and Results. Nine myocardial infarction (MI) patients with clinically documented sustained
monomorphic VT (SMVT), 40 MI patients without clinical VT, and 30 normal healthy control subjects
were evaluated. PVS using three extrastimuli and SAECG recording were performed in the MI patients
on day 36±4 after infarction. Of 40 MI patients, SMVT was inducible in 14, sustained polymorphic VT
in three, nonsustained monomorphic VT in three, nonsustained polymorphic VT in two, and no inducible
arrhythmia was obtained in 18. There were significant differences between MI patients with inducible
SMVT and without inducible SMVT in the following SAECG variables: filtered QRS durations (high-pass
filter setting, 25, 40, and 80 Hz); low-amplitude signal durations (LAS) under 10, 20, 30, and 40 ,uV
(high-pass filter setting, 40 and 80 Hz); root-mean-square voltages (RMS) of the terminal 20, 30, 40, 50,
and 60 msec (high-pass filter setting, 40 and 80 Hz); area ratio (area 20-50 Hz/area 0-20 Hzx 10) of a
120-msec sampling interval starting 20 msec before QRS offset; factor of normality on lead X; and
minimum value of the variables on lead X, Y, or Z. Stepwise logistic regression analysis selected only IAS
under 30 ,tV (high-pass filter setting, 80 Hz) and area ratio as independent predictors of inducible SMVT.
With these two variables, the predicted probability of inducible SMVT [p(VT)] was expressed as
p(VT)=1/[1+exp (6.2-0.11 LAS-0.01 area ratio)]. This function had 93% sensitivity, 81% specificity,
72% positive predictive value, 95% negative predictive value, and 85% predictive accuracy with .0.3 as the
criterion of a positive test.
Conclusions. The combined use of time and frequency domain analysis of SAECG can enhance the
accuracy of this technique as a screening test for results of PVS in MI patients without clinical VT.
(Circulation 1992;86:780-789)
KEY WoRDs * programmed ventricular stimulation
tachycardia * electrocardiography, signal-averaged
Sudden death after myocardial infarction (MI) is
usually caused by ventricular tachyarrhythmias.
Accurate identification of patients prone to ventricular arrhythmias is necessary to treat the group of
patients at high risk rationally. The signal-averaged
From the Second Department of Medicine, Tokyo Medical and
Dental University; The Cardiovascular Division, Tsuchiura Kyodo
General Hospital, Ibaraki, Japan; and the Department of Cardiovascular Diseases, Medical Research Institute, Tokyo Medical and
Dental University.
Supported in part by a grant from Kohokai (A.N.).
Address for reprints: Akihiko Nogami, MD, The Second Department of Medicine, Tokyo Medical and Dental University,
1-5-45 Yushima, Bunkyo-ku, Tokyo 113, Japan.
Received October 8, 1991; revision accepted May 18, 1992.
*
fast Fourier transform analysis
*
ventricular
ECG (SAECG) has been demonstrated to be a useful
noninvasive test for detecting patients at risk for ventricular tachyarrhythmias.1-6 In principle, two methods
are currently used: 1) analysis of the ECG in the time
domain after high-gain amplification and signal averaging1'2 and 2) analysis in the frequency domain by use of
Fourier transformation.3-6 However, both methods
have several limitations. In time domain analysis, highpass filters may disturb the signals, discrimination between noise and late potentials is difficult, the definitions of abnormal findings are inconsistent, and patients
with bundle branch block generally have to be excluded.
In frequency domain analysis, some methodological
prerequisites, such as adequate segment or window
function, have to be considered. Furthermore, the re-
Nogami et al Signal-Averaged ECG in MI
sults of the two methods sometimes vary considerably
even when the same patients are analyzed. The purpose
of this study was to determine whether the combined
use of time and frequency domain variables in SAECG
facilitates identification of patients who will have sustained monomorphic ventricular tachycardia (VT) induced during programmed ventricular stimulation.
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Methods
Patient Population
The study group consisted of 40 consecutive patients
(62±9 years old, 35 men and five women) <75 years of
age admitted to Tsuchiura Kyodo General Hospital
between January and December 1987 because of Q
wave acute MI. Written informed consent was obtained
from all patients undergoing programmed ventricular
stimulation. Patients were excluded from the study if
they had one of the following: uncontrolled angina or
heart failure, sustained VT or ventricular fibrillation, or
significant nonischemic heart disease. Patients with
bundle branch block were not excluded. Nine MI patients (65 ±8 years old, eight men and one woman) with
clinically documented sustained monomorphic VT 1
month after acute MI served as positive control subjects. The normal control group consisted of 30 healthy
volunteers (60±7 years old, 26 men and four women)
without signs of heart disease or hypertension. All
normal control subjects had a normal ECG. The study
protocol was approved by the Institutional Review
Board of our hospital.
Body Surface ECG Recording and Signal Averaging
SAECG recordings were performed on day 36±4
after MI with the Arrhythmia Research Technology
model 1200 EPX (Mortara Instrument, Inc.). No patient received j3-blockers or other antiarrhythmic therapy for at least 7 days before the recording. The body
surface recordings were made during sinus rhythm from
standard bipolar X, Y, and Z leads. At least 300 beats
were averaged during each recording, and in all cases
the noise levels were <0.6 ,uV. After passing through a
template recognition program that rejected ectopic
beats and grossly noisy signals, the signals were initially
filtered (0.05-250 Hz) and amplified.
Time Domain Signal Processing ofAveraged
X, Y, ZLeads
The digitized X, Y, and Z data underwent further
signal processing according to the methods of Simson.2
Specifically, each QRS complex was filtered bidirectionally with a high-pass filter. In all, three high-pass
filterings were used (cutoff, 25, 40, and 80 Hz), with the
low-pass cutoff frequency fixed at 250 Hz. A vector
magnitude was calculated as V=(X2+Y2+Z2)"12 by combining the filtered signals from the three leads. The
onset and offset of the QRS were determined automatically. The following parameters were measured: filtered QRS duration (f-QRS); duration of low-amplitude signals of <10, 20, 30, and 40 ,uV (LAS10, LAS20,
LAS30, and LAS40, respectively); and the root-meansquare voltages of the terminal 10, 20, 30, 40, 50, and 60
msec of the filtered QRS complex (RMS10, RMS20,
RMS30, RMS40, RMS50, and RMS60, respectively).
781
Frequency Domain Signal Processing of Averaged
X, YX Z Leads
Averaged, digitized X, Y, and Z lead data were
transferred to a computer for fast Fourier transformation (FFT).3-6 The 512-point FFTs were calculated by
use of a four-term Blackman-Harris interval to smooth
discontinuities and reduce spectral leakage. We chose
an FF1 sampling interval of 120 msec. To avoid uncertainty about QRS offset, sampling intervals were started
at the QRS onset, 40 msec before QRS offset, 20 msec
before QRS offset, and at the QRS offset. Area ratio
was calculated by dividing the area under the curve
(from the plot of magnitude versus frequency) between
20 and 50 Hz by the area under the curve between 0 and
20 Hz for each of the FFT intervals.34 The area ratios of
the individual X, Y, and Z lead values were averaged
after log transformation. The mean values were expressed as antilogs multiplied by a constant, 1 x 105.
For spectral temporal mapping analysis,5'6 the ST
segment was divided into 25 segments. The first segment
started 20 msec before the QRS offset with a total
segment length of 80 msec, and subsequent segments
started progressively later in the ST segment, at intervals in steps of 2 msec. The frequency components of
each segment were calculated with FFT. The spectra
produced were cross-correlated, and total areas for the
frequency range of 40-140 Hz were produced. From
these two indexes, a single quantitative index, the factor
of normality, was calculated for each lead (X, Y, or Z)
and for the composite lead, i.e., the sum of all the leads.
Ventricular Stimulation Protocol
Programmed ventricular stimulation was performed on
day 36±4 after MI. Stimulation was performed with a
programmable stimulator. The stimuli, with rectangular
pulses 1.0 msec in duration at twice the diastolic threshold,
were introduced from the right ventricular apex and
outflow tract. The following protocol was used: 1) A single
ventricular extrastimulus (S2) was introduced in late diastole during two paced ventricular cycle lengths (S,S =600
msec and 400 msec) and moved earlier until ventricular
refractoriness was encountered. 2) Double ventricular
extrastimuli were then introduced starting at an S1S2
interval 50 msec longer than the ventricular refractory
period and S2S3 equal to S1S2. S2S3 was shortened until S3
failed to capture the ventricle, and S1S2 was then shortened until S3 evoked a response. This was continued until
both S2 and S3 reached refractoriness. 3) A third extrastimulus (S4) was delivered in an analogous fashion during
two paced ventricular cycle lengths. The end point of the
study was either the initiation of a sustained VT or the
completion of the stimulation protocol. The physicians
who performed the programmed ventricular stimulation
did not know the results of the patient's SAECG.
The following definitions were used: 1) Sustained
monomorphic VT is repetitive ventricular response with
uniform QRS configuration lasting .30 seconds or
requiring termination before 30 seconds because of
hemodynamic compromise. 2) Sustained polymorphic
VT is repetitive ventricular response with nonuniform
QRS configuration lasting >30 seconds or requiring
termination before 30 seconds because of hemodynamic
compromise. 3) Ventricular fibrillation is disorganized
ventricular rhythm with nonuniform QRS configuration
782
Circulation Vol 86, No 3 September 1992
TABLE 1. Clinical Characteristics
MI with
clinical
SMVT
(group 1)
9
65+8
MI with
inducible
SMVT
(group 2)
14
63±7
MI without
inducible
SMVT
(group 3)
26
61±9
Control
(group 4)
30
60±7
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
No. of patients
Age (years)
Sex
21
26
14
8
Male
4
5
1
0
Female
Location of infarction
15
6
Anterior
5
2
10
6
Inferior
1
1
1
Anterior and inferior
1
0
1
Other
2
2
2
Two or more infarctions
No. of diseased coronary arteries
6
17
8
One vessel
1
6
4
Two vessels
2
3
2
Three vessels
35+±8*
60±14
41±12t
Ejection fraction (%)
7
7
6
Left ventricular aneurysm
1
4
0
2
Bundle branch block
4
0
4
3
Holter Lown's grade IVb
MI, myocardial infarction; SMVT, sustained monomorphic ventricular tachycardia. Values are
mean±SD.
*p<0.01, tp<0.O5 compared with group 3.
and a mean cycle length <200 msec or no clearly
defined QRS complexes on the surface leads. 4) Nonsustained VT is a series of five or more repetitive
ventricular responses terminating spontaneously in <30
seconds. 5) No inducible arrhythmia is four or fewer
repetitive ventricular responses. The reproducibility of
induced sustained arrhythmias was not checked in this
study.
Cardiac Catheterization
All MI patients underwent complete catheterization
with coronary angiography and left ventriculography.
Significant coronary lesions were defined as .60%
reduction of luminal diameter. Left ventricular ejection
fraction was determined from ventriculography.
Statistics
Continuous values were expressed as mean±+1 SD.
Comparisons of continuous variables were made by
Student's t test after validation of a normal distribution
by the Wilk-Shapiro test.7 Because area ratios and
factors of normality in FFT analysis did not follow a
normal distribution, the variables were expressed as
median with 25th to 75th percentile range and were
examined by a nonparametric method (Mann-Whitney
test).8 Dichotomous variables were analyzed by Fisher's
exact probability test. Multivariate analysis was performed with the logistic multiple regression model with
stepwise variable selection.9 A probability value of
p<O.O5 was accepted as significant.
Results
Results of Programmed Ventricular Stimulation
Of 40 MI patients without clinical VT, programmed
ventricular stimulation induced sustained monomorphic
.
.
.
.
.
.
.
.
.
VT in 14 patients (35%), sustained polymorphic VT in
three (7.5%), nonsustained monomorphic VT in three
(7.5%), nonsustained polymorphic VT in two (5%), and
no inducible arrhythmia in 18 (45%). The mean cycle
length of induced sustained monomorphic VT was
243+±43 msec. Because only induced sustained monomorphic VT has prognostic value, as we have previously
reported,'0'11 patients with other inducible arrhythmias
and patients with no inducible arrhythmia were combined into one group (group 3) for further comparison
with the inducible sustained monomorphic VT group
(group 2). In all nine patients with clinical VT (group
1), sustained monomorphic VT was inducible and the
mean cycle length of induced VT was 280±82 msec.
Clinical characteristics of MI patients with clinical sustained monomorphic VT (group 1), patients with inducible sustained monomorphic VT (group 2), patients
without inducible sustained monomorphic VT (group
3), and normal control subjects (group 4) are summarized in Table 1. Patients in groups 1 and 2 had
significantly lower ejection fraction than those in group
3. There were no significant differences in other variables among groups 1, 2, and 3.
Time Domain Analysis
Results of 33 time domain analysis variables for group
1 (nine patients), group 2 (14 patients), group 3 (26
patients), and group 4 (30 control subjects) are shown in
Table 2. The f-QRS durations were significantly longer
in group 1 than in group 2 or 3, and the f-QRS durations
in group 2 were significantly longer than in group 3 or 4
at all high-pass filter settings. The f-QRS durations in
group 3 were significantly longer than in group 4 at the
40- or 80-Hz high-pass filtering. The LAS durations
Nogami et al Signal-Averaged ECG in MI
783
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
TABLi 2. Time Domain Variables
MI with clinical
MI with inducible
MI without inducible
Variables
SMVT
SMVT
Control
SMVT
(high-pass filter cutoff)
(group 1)
(group 2)
(group 3)
(group 4)
f-QRS (25 Hz) (msec)
155+31
120±31
102±11
97±10
152±29
f-QRS (40 Hz) (msec)
115±21
99±12
92±9
149±32
f-QRS (80 Hz) (msec)
111±22
96+±12
88±9
LAS10 (25 Hz) (msec)
27±19
20±16
14±6
16±6
34+23
LAS20 (25 Hz) (msec)
26±17
19±8
20±6
41±27
LAS30 (25 Hz) (msec)
33±20
24±10
24±7
LAS40 (25 Hz) (msec)
50±32
41±23
28±+11
27+8
LAS10 (40 Hz) (msec)
27±16
19±9
13+6
14±5
LAS20 (40 Hz) (msec)
39±19
31±13
21±10
20±6
LAS30 (40 Hz) (msec)
50±25
39±15
26±11
25±7
54±23
LAS40 (40 Hz) (msec)
44±15
29±12
29±7
37±18
LAS10 (80 Hz) (msec)
27±14
19±6
17±5
LAS20 (80 Hz) (msec)
55±27
43±19
28±11
26±8
61±25
34±11
LAS30 (80 Hz) (msec)
54±22
32±6
74±27
57±21
38±11
LAS40 (80 Hz) (msec)
35±8
5±4
6±7
5±4
3±2
RMS10 (25 Hz) (,uV)
9±11
13±9
13±10
11±8
RMS20 (25 Hz) (,V)
14±15
26±17
RMS30 (25 Hz) (j.V)
19±17
29±20
19±16
RMS40 (25 Hz) (,V)
29±24
54±36
57±39
33±35
RMS5O (25 Hz) (,tV)
61±56
90±57
113±83
52±51
RMS60 (25 Hz) (,uV)
92±71
137±73
162±97
3±3
6±4
3±2
5±4
RMS10 (40 Hz) (,uV)
5±5
7±4
RMS20 (40 Hz) (,V)
13±8
11±6
9±7
12+9
RMS30 (40 Hz) (,V)
26±18
23±13
12±7
18±12
43±28
RMS40 (40 Hz) (,M)
45±25
17±+10
67±39
80±48
33±25
RMS50 (40 Hz) (,M)
26±16
47±33
84±37
RMS60 (40 Hz) (,uV)
108±53
2+2
3±1
3±2
2±2
RMS10 (80Hz) (,V)
3+2
7±4
4±2
7±3
RMS20 (80 Hz) (pV)
6±4
14±7
4±3
12±7
RMS30 (80 Hz) (,uV)
7+4
11±8
21±12
23±10
RMS40 (80 Hz) (,M)
10±9
17±14
30±14
36±14
RMS50 (80 Hz) (pV)
15±11
23+13
33+12
41±16
RMS60 (80 Hz) (,V)
MI, myocardial infarction; SMVT, sustained monomorphic ventricular tachycardia; f-QRS, filtered QRS duration;
LASx, duration of low-amplitude signals of <x ,V; RMSx, root-mean-square voltage of the terminal x msec of the
filtered QRS complex. Values are mean±SD.
were significantly longer in group 1 than in group 3 at
40- or 80-Hz high-pass filtering, and the LAS durations in group 2 at 40- or 80-Hz high-pass filtering
were significantly longer than in group 3 or 4, but there
was no significant difference in the LAS duration
between groups 3 and 4. The RMS voltages of the
terminal 30, 40, 50, or 60 msec of the QRS were
significantly lower in group 1 than in group 3, and the
RMS voltages of the terminal 30, 40, 50, or 60 msec of
the QRS in group 2 at 40- or 80-Hz high-pass filtering
were significantly lower than in group 3 or 4. There
was no significant difference in the RMS duration
between groups 3 and 4.
Frequency Domain Analysis: Area Ratio
FFT area ratios in the four study groups are summarized in Table 3. The results of FFT analysis of SAECG
to detect the low-amplitude, high-frequency signals
were very dependent on the location of the 120-msec
sampling interval for analysis. A significant difference in
area ratios between groups 2 and 3 was found only with
the sampling interval starting 20 msec before QRS
offset. There was no significant difference in area ratio
with this sampling interval between groups 3 and 4.
Area ratios derived from FFT analysis of the remaining
three sampling intervals were not significantly different
between groups 2 and 3.
Frequency Domain Analysis: Factor of Normality
The results of spectral temporal mapping analysis are
summarized in Table 4. Factors of normality on lead X
and the minimum value of the variables on lead X, Y, or
Z in group 1 or 2 were significantly lower than in group
3. The factor of normality on lead X in group 3 was
significantly higher than in group 4.
784
Circulation Vol 86, No 3 September 1992
TABLE 2. Time Domain Variables (cont)
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Variables
(high-pass filter cutoff)
f-QRS (25 Hz) (msec)
f-QRS (40 Hz) (msec)
f-QRS (80 Hz) (msec)
LAS10 (25 Hz) (msec)
LAS20 (25 Hz) (msec)
LAS30 (25 Hz) (msec)
LAS40 (25 Hz) (msec)
LAS10 (40 Hz) (msec)
LAS20 (40 Hz) (msec)
LAS30 (40 Hz) (msec)
LAS40 (40 Hz) (msec)
LAS10 (80 Hz) (msec)
LAS20 (80 Hz) (msec)
LAS30 (80 Hz) (msec)
LAS40 (80 Hz) (msec)
RMS10 (25 Hz) (pAV)
RMS20 (25 Hz) (,V)
RMS30 (25 Hz) (,uV)
RMS40 (25 Hz) (,V)
RMS50 (25 Hz) (,uV)
RMS60 (25 Hz) (,V)
RMS10 (40 Hz) (,uV)
RMS20 (40 Hz) (/.V)
RMS30 (40 Hz) (,uV)
RMS40 (40 Hz) (plV)
RMS50 (40 Hz) (MV)
RMS60 (40 Hz) (,uV)
RMS10 (80 Hz) (,uV)
RMS20 (80 Hz) (pAV)
RMS30 (80 Hz) (,uV)
RMS40 (80 Hz) (lV)
RMS50 (80 Hz) (,V)
RMS60 (80 Hz) (,uV)
Group 1
vs.
group 2
<0.05
<0.01
<0.01
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
<0.05
NS
NS
NS
NS
NS
NS
NS
Group 2
vs.
group 3
<0.01
<0.01
<0.05
NS
NS
NS
NS
<0.05
<0.05
<0.01
<0.01
<0.05
<0.05
<0.01
<0.01
NS
NS
NS
<0.05
NS
NS
NS
<0.01
<0.01
<0.01
<0.01
<0.01
NS
<0.05
<0.01
<0.01
<0.01
<0.05
Multivariate Analysis
All 26 SAECG variables (the 22 time domain and the
four frequency domain variables) in which there were
significant differences between groups 2 and 3 were
chosen as candidate variables in a multivariate analysis.
Stepwise logistic regression analysis revealed that
LAS30 at 80-Hz high-pass filtering (p<0.05) and area
ratio with the sampling interval starting 20 msec before
QRS offset (p<0.05) were the only independent predictors of inducible sustained monomorphic VT in
patients with MI (Table 5). With this model, the predicted probability of inducible sustained monomorphic
VT [p(VT)] was expressed as
p(VT)= 1/(1 +elgt)
where logit=6.2-0.11(LAS30)-0.01(area ratio).
Optimal Cutoff Values
To determine the optimal cutoff points for variables,
receiver operating characteristic (ROC) curve analysis
p
Group 3
vs.
group 4
NS
<0.05
<0.01
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
Group 1
vs.
group 3
<0.01
<0.01
<0.01
NS
NS
NS
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.05
<0.01
NS
NS
NS
<0.01
<0.01
<0.01
<0.05
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
Group 2
vs.
group 4
<0.01
<0.01
<0.01
NS
NS
<0.05
<0.01
<0.05
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
<0.01
NS
NS
NS
<0.05
<0.05
<0.05
NS
NS
<0.01
<0.01
<0.01
<0.01
NS
NS
<0.05
<0.01
<0.01
<0.01
was performed (Figure 1). ROC curves were plotted for
LAS30 at 80-Hz high-pass filtering, area ratio with the
sampling interval starting 20 msec before QRS offset,
and p(VT), which was calculated from LAS30 and area
ratio with a logistic regression model. These were
inspected for the cutoff point, which mammized a
true-positive ratio while maintaining a false-positive
ratio at c50%. These high-sensitivity, moderate-specificity points were chosen because it is desirable to have
a screening test that does not produce many falsenegative results. ROC curve analysis identified highsensitivity cutoff criteria for LAS30 .37 msec, area
ratio 2101, and p(VT) >0.3.
Individual patient values for LAS30, area ratio, and
p(VT) are plotted in Figure 2. Although there were
significant differences in the LAS30 and the area ratio
between groups 2 and 3, there were many false-positive
and false-negative results. LAS30 had 86% sensitivity,
62% specificity, 55% positive predictive value, 89%
negative predictive value, and 70% predictive accuracy.
Nogami et al Signal-Averaged ECG in MI
785
TABLE 3. Frequency Domain Variables 1: Area Ratio
P
MI with MI without
inducible
Group 1 Group 2 Group 3 Group 1 Group 2
inducible
vs.
vs.
SMVT
Control
vs.
vs.
vs.
Sampling interval
SMVT
starting point
(group 1) (group 2) (group 3) (group 4) group 2 group 3 group 4 group 3 group 4
NS
NS
693
657
NS
NS
NS
QRS onset
605
627
(546, 774) (598, 837) (624, 960) (512, 745)
NS
NS
40 msec before QRS offset
237
223
130
192
NS
NS
NS
(76, 383)
(84, 650) (103, 405) (90, 264)
NS
<0.01
20 msec before QRS offset
97
120
87
57
NS
<0.01
NS
(41, 124)
(39, 84)
(65, 188) (55, 210)
NS
<0.01
110
61
46
NS
NS
<0.01
QRS offset
96
(47, 155)
(43, 50)
(70, 146) (65, 138)
MI, myocardial infarction; SMVT, sustained monomorphic ventricular tachycardia. Values are medians; 25th to 75th percentile range in
parentheses.
Area ratio: 20-50 Hz/area 0-20 Hzx1io.
MI with
clinical
SMVT
Comparison With Other Determinants of Inducibility
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
The area ratio had values of 64%, 65%, 50%, 77%, and
65%, respectively. On the other hand, p(VT) showed
fewer false-positive and false-negative results and had
93% sensitivity, 81% specificity, 72% positive predictive
value, 95% negative predictive value, and 85% predictive accuracy using p(VT) >0.3 as the criterion of a
positive test. Also, p(VT) determined all nine group 1
patients as positive and 29 of 30 group 4 patients as
negative.
In the present study, another variable associated with
inducible sustained monomorphic VT was depressed
left ventricular ejection fraction (Table 1). To assess the
relative importance of left ventricular ejection fraction
and SAECG variables (LAS30 at 80-Hz high-pass filtering and area ratio), a stepwise logistic regression
analysis was performed with these three variables as the
potential predictors of inducibility. Stepwise logistic
regression analysis selected only LAS30 and area ratio
as independent predictors of inducible sustained monomorphic VT. Left ventricular ejection fraction was the
least powerful predictor of inducibility among these
three variables.
Subgroups of Group 2
MI patients with inducible sustained monomorphic
VT (group 2) were divided into subgroups according to
the cycle length of induced VT and the number of
extrastimuli required to initiate VT (Table 6). There
was no significant difference in SAECG variables or
p(VT) between patients in whom one or two extrastimuli were required to initiate VT and patients in whom
three extrastimuli were needed. When two subgroups
were made according to the cycle length of induced VT
(cycle length .250 msec or <250 msec), there were
significant differences in several time domain variables
between the two subgroups. Patients with the induced
VT cycle length .250 msec had significantly longer
LAS40 at 25- or 40-Hz high-pass filtering and lower
RMS voltages of the terminal 40 and 60 msec of the
QRS at 25-, 40-, or 80-Hz high-pass filtering than
patients with the induced VT cycle length <250 msec.
There was no significant difference in frequency domain
variables (area ratios or factors of normality) between
patients with VT cycle length .250 msec and <250
Discussion
In the present study, we tested the hypothesis that the
combined use of time and frequency domain variables
of SAECG would improve the selection of MI patients
with inducible sustained monomorphic ventricular
tachycardias by programmed ventricular stimulation.
Improved Selection of MI Patients for Programmed
Ventricular Stimulation
Several investigators10-18 have reported that programmed ventricular stimulation is useful to identify
patients with MI at risk for ventricular tachyarrhythmias
or sudden cardiac death. However, it is not a desirable
screening test because of the invasive nature of programmed ventricular stimulation. The incidence of inducible sustained monomorphic VT has been reported
to be 7-25%. In view of the expense, low incidence of
msec.
TABLE 4. Frequency Domain Variables 2: Factor of Normality
MI with
MI with
MI without
clinical
SMVT
inducible
SMVT
inducible
SMVT
(group 3)
90±15
(%Mo)
P
Control
Group 1
vs.
group 2
Group 2
vs.
group 3
Group 3
vs.
group 4
Lead
(group 4)
(group 1) (group 2)
<0.01
<0.05
<0.05
75±27
43±24
69+29
Lead X
NS
<0.05
72±27
NS
72±28
46±25
53±29
Lead Y
NS
NS
73±24
NS
76±29
62±31
68±33
Lead Z
NS
<0.05
NS
56±23
44±29
60±26
29±15
Minimum (X, Y, Z)
NS
NS
NS
72±27
63±29
51±20
55±19
Composite
MI, myocardial infarction; SMVT, sustained monomorphic ventricular tachycardia. Values are mean±SD.
Group 1
vs.
group 3
<0.01
<0.05
NS
<0.01
NS
Group 2
vs.
group 4
NS
<0.05
NS
NS
<0.05
786
Circulation Vol 86, No 3 September 1992
TABLE 5. Multivariate Analysis: Stepwise Logistic Regression
Analysis
Variable
Coefficient
p
LAS30 (high-pass filter setting, 80 Hz)
<0.05
-0.11
Area ratio
<0.05
-0.01
(starting point, 20 msec before QRS offset)
Constant
6.2
LAS30, duration of low-amplitude signals of <30 ,V.
induction of sustained monomorphic VT, patients' psychological stress, and the potential morbidity of electrophysiological testing, improved noninvasive methods for
selecting patients for programmed ventricular stimulation are needed. The present study reveals that the
combined use of time and frequency domain analysis of
SAECG is the most accurate predictor of inducible
sustained monomorphic VT.
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Previous Studies
Many investigators'9-23 have reported a strong correlation between SAECG variables and inducibility of
sustained monomorphic VT. However, the results were
limited by the patient selection bias that existed in most
of these studies. In most previous studies,19-21,23 the
patients evaluated included those presenting with sustained VT, nonsustained VT, or syncope. We believe
that to determine the usefulness of SAECG as a screening test for programmed ventricular stimulation after
MI, the study population should be patients without
clinical ventricular tachyarrhythmias. Patients with clinical ventricular tachyarrhythmias do not require screening tests because they are already known to be candidates for programmed ventricular stimulation. In the
present study, MI patients without clinical sustained
ventricular tachyarrhythmias were analyzed to determine the predictors of inducible sustained monomorphic VT. Denniss et a122 evaluated 306 patients with
recent MI and no spontaneous ventricular tachyarrhythmias and reported that as a predictor of the presence of
inducible sustained VT, QRS duration >140 msec had
a sensitivity of 72%, a specificity of 78%, and a predictive accuracy of 76%. However, they examined only
QRS duration, and their programmed ventricular stimulation protocol was limited to use of two extrastimuli.
Enhancing the Predictive Accuracy by Combining
the SAECG Variables
In previous reports, SAECG parameters of either
time domain or frequency domain were used singly to
predict inducibility, and predictive accuracies were calculated by single cutoff values. In general, however, it is
difficult to determine single cutoff values because of
numerous crossover data points. By using generous
criteria of one variable, one would increase sensitivity,
but specificity would decrease, and by strict criteria,
specificity would increase but sensitivity would decrease. In the present study, the combined use of LAS30
(high-pass filter setting, 80 Hz) and area ratio (area
20-50 Hz/area 0-20 Hzx 10') reduced the false-positive
and false-negative results and enhanced predictive accuracy. Concerning the filter settings, Caref et a123
reported that the combination of time domain variables
A
LAS 30 (80Hz)
Ai ,
.0-1.1
.
O.9
.
37
40
0.8
_ 0.7
a 0.6
-
0.5
7
i 0.3
0.2
0.1
0 0.1 0.2 0.3 04 0.5 0U6 0.7 U8 0.9 1.10
FALSE-POSITIVE RATI0
B
AREA RATIO
1.0
0.9- -
-
52
o
or
0.4-- 190
0.50.6-
2
B-
C
-
30
0.3
0.2
350
0.1
0
0 0.1 0.2 0.3 0.4 0.5 0.8 0.7 U 0.9 1.1 ,o
FALSE-POSITIVE RATIO
P [VT]
X
E
We
$4
0.3 0.4 0.5 0.6 0.7 0.8 0.9
FALSE-OSITIVE RATIO
FIGURE 1. Receiver operating characteristic (ROC) curves
for the three signal-averaged ECG measurements with the
strongest relation to outcome of electrophysiological study:
duration of low-amplitude signals <30 gVat 80-Hz high-pass
filtering (LAS30 [80Hz], panel A), area ratio with the
sampling interval starting 20 msec before QRS offset (panel
B), and predicted probabilities of inducible sustained monomorphic ventricular tachycardia (p[VT]) determined by a
logistic regression model (panel C). Cutoff values were selected for each measurement by inspecting the curve and
identifying the point at which true-positive ratio is maximized
while maintaining a false-positive ratio c50%. Arrows indicate thesepoints on ROC curves. Values for measurements are
indicated along ROC curves.
Nogami et al Signal-Averaged ECG in MI
A ,
120
A2
<"u
-
C
00
0n
0
r <UlDt
r -
Iw
1Dr
787
U[l-
*0
0
D:
00
0
00
A
al F
0
00
0
00
o0
a000
00
00
40
I
.a-
Lat
A
OA_
£
201-
0
L~~
00
,,0
A-
20
~
t3
mRI GoRWm
-
00
ORWI R
WIV
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
B
--
-o0
0
LA
AL.
0.2
~~~~~~~ALA
-a
Ow
00
o00
LA
&
AflGL4£,&
09
080
000
IRGOWI
FIGURE 2. Panel A: Individual data points for duration of
low-amplitude signals <30 gV at 80 Hz high-pass filtering
(LAS30 [80Hz]). Dashed line indicates cutoff criterion identified by receiver operating characteristic (ROC) analysis.
Vertical bars represent 1 SD. Panel B: Individual data points
for area ratio with the sampling interval starting 20 msec
before QRS offset. Values are plotted along the log scale (y
axis). Dashed line indicates cutoff criterion identified by ROC
analysis. Box shows the median and the 25th and 75th
quartiles for the data, and whiskers show the range of the data.
Panel C: Predicted probabilities ofinducible sustained monomorphic ventricular tachycardia (P[VT1) determined by a
logistic regression model. Dashed line indicates cutoff criterion identified by ROC analysis.
'I
GROWU
GROUPM RPIV
analyzed at different filter settings enhanced the
of the technique.
accu-
racy
Cycle Length of Induced VT
Interestingly, there were significant differences in
time domain variables between patients with inducible
slower VT (cycle length 2250 msec) and faster VT
(cycle length <250 msec) but no significant differences
in the frequency domain variables (area ratios or factors
of normality). Because the inducible slower monomorphic VT appears to have more sinister prognostic
implications,10l11l7,18 time domain variables may be
more useful predictors than frequency domain variables
from this point of view. Multivariate analysis to find
independent predictors of inducible slow sustained VT
could not be performed in this study because of the
small number of patients with inducible slow VT (n =6).
Although it did not reach a significant level, p(VT) was
higher in the slow VT group (0.76±0.24) than in the fast
VT group (0.52±0.29).
MI Patients With Clinical VT and Normal
Control Subjects
In the present study, MI patients with clinically documented sustained monomorphic VT (group 1) as positive
controls and normal control subjects (group 4) were also
evaluated. The p(VT) function determined all group 1
patients as positive and 29 of 30 group 4 subjects as
negative, although there were many false-positives (positives in group 4) and false-negatives (negatives in group 1)
when a single SAECG variable was used. The patients
with clinical VT were already known to be candidates for
programmed ventricular stimulation. However, this demonstrates the accuracy of this function.
Implications for Clinical Management
The results of this study indicate that SAECG can be
applied in patients after MI with high sensitivity and
high negative predictive value for inducible sustained
monomorphic VT. This is ideal in this clinical circumstance because it identifies a large group of patients who
have a low probability of inducible VT by programmed
ventricular stimulation. Also identified is a second
group with a higher probability of inducible VT who can
be evaluated more fully with programmed stimulation in
the hope of selecting those who would benefit from
antiarrhythmic treatment.
Limitations
To determine the significance of SAECG for prediction of inducible VT by programmed ventricular stim-
788
Circulation Vol 86, No 3 September 1992
TABLE 6. Comparison Between Subgroups of Group 2
Inducible SMVT
CL<250 msec
CL2250 msec
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Variable
f-QRS (25 Hz) (msec)
f-QRS (40 Hz) (msec)
f-QRS (80 Hz) (msec)
LAS40 (25 Hz) (msec)
LAS40 (40 Hz) (msec)
LAS30 (80 Hz) (msec)
LAS40 (80 Hz) (msec)
RMS40 (25 Hz) (,uV)
RMS60 (25 Hz) (I.V)
RMS40 (40 Hz) (,V)
RMS60 (40 Hz) (,V)
RMS40 (80 Hz) (gV)
RMS60 (80 Hz) (,rV)
Area ratio
ES=1, 2
Inducible SMVT
ES=3
(n=6)
(n=8)
p
(n=5)
(n=9)
136±40
125±27
121±28
58±25
55±14
61±22
63±21
12±8
32±26
11±10
24±13
6±3
15±7
145
108±13
108±14
104±15
28±9
36±9
48±23
52±21
43±23
137±59
23±11
65±32
14±8
29±14
120
NS
NS
NS
0.05
0.01
NS
NS
<0.01
<0.01
<0.05
<0.05
<0.05
<0.05
NS
134±49
124±34
120±36
48±36
45±20
55±28
62±23
30±31
76±69
19±16
40±34
14±8
22±13
102
(55, 210)
(78, 278)
58±28
55±25
69+32
112±9
110±9
107±10
37±12
43±12
53±20
53±20
29±21
101±74
17±9
51±33
11±9
24±14
132
(54, 334)
64±32
57±29
65±32
46±31
51±17
0.63±0.31
p
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
NS
(67, 205)
NS
79±23
NS
NS
45±31
NS
NS
72±37
NS
42+23
NS
41±27
NS
53+23
NS
63±21
NS
0.52±0.29
NS
0.60±0.27
NS
ventricular tachycardia; CL, cycle length; ES, number of extrastimuli required to initiate ventricular
tachycardia; f-QRS, filtered QRS duration; LASx, duration of low-amplitude signals of <x ,uV; RMSx, root-mean-square voltage of the
terminal x msec of the filtered QRS complex; NF, factor of normality; p(VT), predicted probability of inducible sustained monomorphic
ventricular tachycardia. Values except area ratios are mean±SD. Area ratios are medians; 25th to 75th percentile range in parentheses.
NF (lead X)
NF (lead Y)
NF (lead Z)
NF (minimum X, Y, Z)
NF (composite)
p(VT)
SMVT, sustained monomorphic
84±27
49±36
66+37
48+37
59±12
0.76±0.24
ulation, the following need to be taken into account: 1)
changes in the inducibility of VT after MI,11 2) changes
in the variables of SAECG after MI,24 and 3) the
relation between SAECG parameters and the site of the
MI.25 When the inducibility of VT and SAECG varies
over time, the prediction of inducibility should also be
considered to be time dependent. In the present study,
only the inducibility of sustained monomorphic VT on
day 36 after MI was evaluated by SAECG on day 36
after MI. The presence of an abnormal SAECG is
strongly determined by the site of the MI. Nearly twice
as many patients with inferior wall MI have abnormal
SAECG parameters as patients with anterior wall MI.25
Prediction of inducible VT should be studied in a large
number of patients with each type of MI.
In the present study, the incidence of inducible
sustained monomorphic VT was higher (35%) than in
previous studies. Up to three extrastimuli during two
paced ventricular cycle lengths were used to initiate VT,
and all sustained VTs with any cycle lengths were
included in group 2. Reproducibility of VT was not
tested in this study. However, a higher incidence of
inducibility in this study might be a result of the smaller
sample size. The incidence of inducible sustained monomorphic VT in our hospitals was reported to be 23-25%
in a large number of patients when the same ventricular
stimulation protocol was used.10"
Conclusions
In summary, the combined use of frequency domain
and time domain analysis of SAECG can enhance the
accuracy of this technique as a screening test for the
results of programmed ventricular stimulation in MI
patients without clinical ventricular tachyarrhythmias.
Acknowledgment
We thank John J. Rozanski, MD, Cleveland Clinic Florida,
Ft. Lauderdale, Fla., for his helpful discussion.
References
1. Rozanski JJ, Mortara D, Myerburg RJ, Castellanos A: Body surface detection of delayed depolarizations in patients with recurrent
ventricular tachycardia and left ventricular aneurysm. Circulation
1981;63:1172-1178
2. Simson MB: Use of signals in the terminal QRS complex to identify patients with ventricular tachycardia after myocardial infarction. Circulation 1986;74:731-745
3. Cain ME, Ambos HD, Witkowski FX, Sobel BE: Fast-Fourier
transform analysis of signal-averaged electrocardiograms for identification of patients prone to sustained ventricular tachycardia.
Circulation 1984;69:711-720
4. Cain ME, Ambos HD, Markham J, Fischer AE, Sobel BE: Quantification of differences in frequency content of signal-averaged
electrocardiograms in patients with compared to those without
sustained ventricular tachycardia. Am J Cardiol 1985;55:1500-1505
5. Haberl R, Jilge G, Pulter R, Steinbeck G: Comparison of frequency and time domain analysis of the signal-averaged electrocardiogram in patients with ventricular tachycardia and coronary
artery disease: Methodologic validation and clinical relevance.
JAm Coll Cardiol 1988;12:150-158
6. Haberl R, Jilge G, Pulter R, Steinbeck G: Spectral mapping of the
electrocardiogram with Fourier transform for identification of
patients with sustained ventricular tachycardia and coronary artery
disease. Eur Heart J 1989;10:316-322
7. Shapiro SS, Wilk MB: An analysis variance test for normality
(complete samples). Biometrika 1965;52:591-611
8. Armitage P: Distribution-free method, in: Statistical Methods in
Medical Research. Oxford, Blackwell, 1971, pp 394-407
Nogami et al Signal-Averaged ECG in MI
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
9. Armitage P: Further analysis of qualitative data, in: Statistical
Methods in Medical Research. Oxford, Blackwell, 1971, pp 362-393
10. Iesaka Y, Nogami A, Aonuma K, Nitta J, Chun YH, Fujiwara H,
Hiraoka M: Prognostic significance of sustained monomorphic ventricular tachycardia induced by programmed ventricular stimulation using up to triple extra-stimuli in survivors of acute myocardial
infarction. Am J Cardiol 1990;65:1057-1063
11. Nogami A, Aonuma K, Takahashi A, Nitta J, Chun YH, Iesaka Y,
Hiroe M, Marumo F: Usefulness of early versus late programmed
ventricular stimulation in acute myocardial infarction. Am J Cardiol 1991;68:13-20
12. Green HL, Reid PR, Schaffer AH: The repetitive ventricular
response in man: A predictor of sudden death. N Engl J Med
1978;298:729-734
13. Hamer A, Vohra J, Hunt D, Sloman G: Prediction of sudden death
by electrophysiologic studies in high risk patients surviving acute
myocardial infarction. Am J Cardiol 1982;50:223-229
14. Richards DA, Cody DV, Denniss AR, Russell PA, Young AA,
Uther JB: Ventricular electrical instability: A predictor of death
after myocardial infarction. Am J Cardiol 1983;51:75-80
15. Waspe LE, Seinfeld D, Ferrick A, Kim SG, Matos JA, Fisher JD:
Prediction of sudden death and spontaneous ventricular tachycardia in survivors of complicated myocardial infarction: Value of the
response to programmed stimulation using a maximum of three
ventricular extrastimuli. JAm Coll Cardiol 1985;5:1292-1301
16. Cripps T, Bennett ED, Camm AJ, Ward DE: Inducibility of sustained monomorphic ventricular tachycardia as a prognostic indicator in survivors of recent myocardial infarction: A prospective
evaluation in relation to other prognostic variables. J Am Coll
Cardiol 1989;14:289-296
17. Breithardt G, Borggrefe M, Haerten K: Role of programmed ventricular stimulation and noninvasive recording of ventricular late
potentials for the identification of patients at risk of ventricular
tachyarrhythmias after myocardial infarction, in Zipes DP, Jalife J
(eds): Cardiac Electrophysiology and Arrhythmias. New York,
Grune and Stratton, 1985, pp 553-561
789
18. Denniss AR, Richards DA, Cody DV, Russell 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
electrocardiograms of survivors of acute myocardial infarction.
Circulation 1986;74:731-745
19. Kanovsky MS, Falcone RA, Dresden CA, Josephson ME, Simson
MB: Identification of patients with ventricular tachycardia after
myocardial infarction: Signal-averaged electrocardiogram, Holter
monitoring, and cardiac catheterization. Circulation 1984;70:
264-270
20. Nalos PC, Gang ES, Mandel WJ, Landenheim ML, Lass Y, Peter
T: The signal-averaged electrocardiogram as a screening test for
inducibility of sustained ventricular tachycardia in high risk
patients: A prospective study. Am J Coll Cardiol 1987;9:539-548
21. Lindsay BD, Ambos HD, Schechtman KB, Cain ME: Improved
selection of patients for programmed ventricular stimulation by
frequency analysis of signal-averaged electrocardiograms. Circulation 1986;73:675-683
22. Denniss AR, Richards DA, Cody DV, Russell PA, Young AA,
Ross DL, Uther JB: Correlation between signal-averaged electrocardiogram and programmed stimulation in patients with and without spontaneous ventricular tachyarrhythmias. Am J Cardiol 1987;
59:586-590
23. Caref EB, Turitto G, Ibrahim BB, Henkin R, El-Sherif N: Role of
bandpass filters in optimizing the value of the signal-averaged
electrocardiogram as a predictor of the results of programmed
stimulation. Am J Cardiol 1989;64:16-26
24. Kuchar DL, Thorburn CW, Sammel NL: Late potentials detected
after myocardial infarction: Natural history and prognostic significance. Circulation 1986;74:1280-1289
25. Gomes JA, Winters SL, Martinson M, Machac J, Stewart D, Targonski A: The prognostic significance of quantitative signalaveraged variables relative to clinical variables, site of myocardial
infarction, ejection fraction and ventricular premature beats: A
prospective study. JAm Coll Cardiol 1989;13:377-384
Combined use of time and frequency domain variables in signal-averaged ECG as a
predictor of inducible sustained monomorphic ventricular tachycardia in myocardial
infarction.
A Nogami, Y Iesaka, J Akiyama, A Takahashi, J Nitta, Y Chun, K Aonuma, M Hiroe, F
Marumo and M Hiraoka
Downloaded from http://circ.ahajournals.org/ by guest on June 14, 2017
Circulation. 1992;86:780-789
doi: 10.1161/01.CIR.86.3.780
Circulation is published by the American Heart Association, 7272 Greenville Avenue, Dallas, TX 75231
Copyright © 1992 American Heart Association, Inc. All rights reserved.
Print ISSN: 0009-7322. Online ISSN: 1524-4539
The online version of this article, along with updated information and services, is located on
the World Wide Web at:
http://circ.ahajournals.org/content/86/3/780
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally
published in Circulation can be obtained via RightsLink, a service of the Copyright Clearance Center, not the
Editorial Office. Once the online version of the published article for which permission is being requested is
located, click Request Permissions in the middle column of the Web page under Services. Further
information about this process is available in the Permissions and Rights Question and Answer document.
Reprints: Information about reprints can be found online at:
http://www.lww.com/reprints
Subscriptions: Information about subscribing to Circulation is online at:
http://circ.ahajournals.org//subscriptions/