Mapping of Ventricular Repolarization Potentials in Patients With Arrhythmogenic Right Ventricular Dysplasia
Principal Component Analysis of the ST-T Waves
Background Nonuniform recovery of ventricular excitability has been demonstrated to facilitate the reentry circuits leading to the development of ventricular tachyarrhythmias. This can also occur in arrhythmogenic right ventricular dysplasia (ARVD). In fact, in patients with ARVD, abnormalities of ventricular repolarization are often observed on 12-lead ECGs, but their predictive value for the occurrence of malignant arrhythmias is yet to be established. Because body-surface potential mapping has been proved to be useful for the detection of heterogeneities in ventricular recovery even though they are not revealed by conventional 12-lead ECGs, we attempted to analyze repolarization potentials on the entire chest surface to find abnormalities that can be predictive of ventricular arrhythmias.
Methods and Results Body-surface potential maps were recorded from 62 anterior and posterior thoracic leads in 22 patients affected by ARVD, 9 with episodes of sustained ventricular tachycardias (VT) and 13 without. Thirty-five healthy subjects were also studied as control subjects. The 62 chest ECGs were simultaneously recorded, digitally converted at a rate of 2000 Hz, and stored on a hard disk of a body-surface mapping computer system. In each subject, the QRST integral map was obtained by calculating at each lead point the algebraic sum of all instantaneous potentials, from the QRS onset to the T-wave end, multiplied by the sampling interval. In most ARVD patients, we observed a larger-than-normal area of negative values on the right anterior thorax. This abnormal pattern could be explained by a delayed repolarization of the right ventricle. Nevertheless, it was not related to the occurrence of VT in our patient population. To detect minor heterogeneities of ventricular repolarization, the principal component analysis was applied to the 62 ST-T waves recorded in each subject. We assumed that a low value of the first or of the first three components (components 1, 2, and 3) indicates a greater-than-normal variety of the ST-T waves, a likely expression of a more complex recovery process. The mean values of the first three components were not significantly different in ARVD patients and control subjects. Nevertheless, considering the two subsets of patients with and without VT, the values of component 1, components 1+2, and component 1+2+3 were significantly lower in the group of ARVD patients with VT. Values of component 1 < 69% (equal to 1 SD below the mean value for control subjects) were found in 6 of 9 VT patients and in 1 patient without VT (sensitivity, 67%; specificity, 92%). A low value of component 1 was the only variable significantly associated with the occurrence of VT.
Conclusions Principal component analysis provides a better quantitative assessment of the complexity of repolarization than other ECG measurements. When applied to ARVD patients, principal component analysis of the ST-T waves recorded from the entire chest surface revealed abnormalities not detected by conventional ECG that can be considered indexes of arrhythmia vulnerability.
Although different factors can play a role in the genesis of arrhythmias, nonuniform recovery of ventricular excitability has been demonstrated to facilitate the reentry circuits leading to the development of ventricular tachyarrhythmias.1,2 This can also occur in arrhythmogenic right ventricular dysplasia (ARVD), a cardiomyopathy associated with frequent occurrence of ventricular arrhythmias.3 In fact, in patients with ARVD, abnormalities of ventricular repolarization are often observed on 12-lead ECGs as negative T waves in the precordial leads beyond V2.3–5 However, the importance of these abnormalities in the actual occurrence of malignant arrhythmias and their predictive value are yet to be established.
Because body-surface potential mapping has been proved to be a useful method for detecting heterogeneities of ventricular recovery even though they are not revealed by conventional 12-lead ECGs,6–10 we attempted to analyze repolarization potentials on the entire chest surface to find abnormalities that can be predictive of ventricular arrhythmias.
We studied 22 patients affected by ARVD, 15 men and 7 women who were 18 to 66 years of age (mean age, 42±13 years) [Table 1⇓]. The diagnosis of ARVD was based on ECG, echocardiographic, and cardiac magnetic resonance findings, according to the recommendations of the Task Force of the European Society of Cardiology.11 Patients with right bundle-branch block (QRS duration >120 ms) were excluded. Nine patients had experienced episodes of sustained ventricular tachycardia (VT), whereas 13 presented only frequent premature ventricular beats or short runs of nonsustained VT (4 patients) with a left bundle-branch block pattern of QRS. No patient had symptoms or signs of right ventricular failure. At the time of body-surface potential mapping recording, among patients with VT, 4 were on amiodarone, 1 was on propafenone, and 1 was taking sotalol. Among patients without VT, 1 was taking flecainide and 1 was on propafenone.
Thirty-five healthy subjects, 26 men and 9 women who were 14 to 54 years of age (mean, 33±10 years), were also studied as control subjects. Informed consent was given by all subjects.
A signal-averaged ECG was recorded by use of a Marquette MAC15 HiRes recorder. Then 250 beats (X, Y, Z Frank leads) were averaged and filtered at 40 to 250 Hz. At least two of the following criteria were required to define the presence of ventricular late potentials: filtered QRS duration >120 ms, RMS voltage of terminal 40 ms <20 μV, and duration of high-frequency low-amplitude signals >38 ms.
Body-Surface Potential Recording
By means of vertical rubber straps, 62 electrodes were applied to the anterior and posterior chest. We adopted the same lead array configuration used by SippensGroenewegen et al.12 Maps were obtained by use of a portable mapping system consisting of a 62-lead electrode set, a front-end (containing 64 ECG preamplifiers and a 16-bit AD converter), and a 80486 PC for data acquisition and processing. Wilson’s central terminal was taken as the reference point for measuring chest potentials. The 62 unipolar ECGs were simultaneously recorded for 10 to 20 seconds and digitally converted at 2000 Hz. Digitized data were sent from the front end to the computer through a fiber-optic transmitter, stored on a hard disk, and then processed. Different kinds of analysis were performed.
QRST Integral Map
The QRST integral maps were obtained by calculating at each lead point the algebraic sum of all instantaneous potentials from the QRS onset to the T wave end multiplied by the sampling interval. The values in millivolts times milliseconds were plotted on the diagram representing the thoracic surface, and isointegral contour lines were drawn.
The QRST mean maps for ARVD and control groups were calculated by use of the respective mean values at each lead point. The mean difference map was obtained by subtracting the mean map of the control subjects from that of the ARVD patients. The mean QRST map of the 35 normal subjects was subtracted from the integral map of each patient, yielding individual difference maps. The resulting value at each lead point was then divided by the SD of the normal values for that point, thus giving the standardized differences from normal values (deviation index map). We considered “abnormal” any area comprising at least three adjacent lead points at which the integral values differed ≥2 SD from mean normal values.
In each subject, the QT interval duration of the beat selected for mapping was measured on the root mean square plot of the 62 leads. Bazett’s formula was applied for calculating QTc values by use of the preceding R-R interval.
Principal Component Analysis
This technique was originally used with the aim of representing a large number of ECG tracings from the thoracic surface by means of a limited number of fundamental waveforms.13 In this study, we applied principal component analysis to all the recorded ST-T waves in each subject studied.
The ST-T waves were divided into successive 20-ms intervals, and the mean potential value of each interval was considered, thus minimizing the noise and the 50-Hz mains interference. The ST-T waves are thus represented by a discrete series of 15 to 20 values. Principal component analysis allows the identification of one set of values corresponding to the “first principal component,” which better represents, by means of appropriate multiplication factors, most of the original sets of data recorded. Usually, the first three components provide nearly the total variation (ie, information) of the original data and the first component gives the greatest contribution. A low value of the first or of the first three components (components 1, 2, and 3), corresponding to a high relative value of the higher-order components, indicates a large variety of the ST-T waves, a likely expression of a more complex recovery process.
Data were presented as mean±SD. Comparisons of sample mean values were made by means of the t test for unpaired data. Multiple regression analysis was used for comparisons among continuous variables. Logistic regression analysis14 was performed to estimate the association between categorical variables and the occurrence of VT by use of the SPSS package. Principal component analysis of the ST-T waves was performed with the “data reduction and factor analysis” procedure contained in the SPSS package. Values of P<.05 were considered significant.
In all control subjects, the QRST integral maps showed a bipolar distribution of the values with a minimum in the upper sternal–right clavicular region and a maximum in the left mammary–axillary region. The mean map (Fig 1A⇓) reproduces this pattern typical for normal subjects already reported by different authors.7,8,15,16
The mean QRST integral map of ARVD patients showed a bipolar distribution similar to that of normal subjects but with a larger negative area on the right anterior thorax (Fig 1B⇑). This was better evidenced by the difference map showing a broad negative area covering nearly all the anterior thorax and the lower portions of the trunk (Fig 1C⇑).
The average values of the QRST maximum and minimum in the control subjects and ARVD patients were 105±42 versus 80±45 mVms (P<.05) and −39±18 versus −37±17 mVms (P= NS), respectively.
The mean values of the first three components were not significantly different in ARVD patients and in control subjects, whereas the sums of the first two and of the first three component were significantly higher in normal subjects (Table 2⇓). Considering the two subsets of patients with and without VT, the values of component 1, of components 1+2, and of components 1+2+3 were found to be significantly lower in the group of ARVD patients with VT compared with patients without VT (Table 2⇓). Taking a value of 69% as a cutoff level for component1 (equal to 1 SD below the mean value for control subjects), we found a lower value in 6 of 9 VT patients and in 1 patient without VT that gave a sensitivity of 67%, a specificity of 92%, and a positive predictive value of 86% for occurrence of VT.
This study demonstrates that abnormalities of ventricular repolarization are frequently observed in patients with ARVD. Specifically, ARVD tends to produce heterogeneities in repolarization that seem to be correlated with vulnerability to malignant ventricular arrhythmias.
The QRST integral maps contain valuable information on the ventricular recovery process.17,18 They mainly reflect the intrinsic repolarization properties and are largely independent of ventricular excitation sequence.18,19 Actually, at the chest surface, negative QRST integrals are recorded from areas facing myocardial regions with longer recovery durations, whereas positive values are recorded from the thoracic surface facing cardiac regions with shorter recovery durations.
In most ARVD patients, we observed a larger-than-normal area of negative values on the right anterior thorax. This pattern could be explained by delayed repolarization of the right ventricle, which most likely is due to myocardial involvement with fibrolipomatous infiltrations of the myocardium and possibly to dilatation of the cavity of various degrees. Nevertheless, this abnormal pattern seems not to be related to the occurrence of VT in our patient population. In fact, only 5 of 9 patients presenting an area of negative values 2 SD lower than normal had episodes of VT.
Heterogeneities of the ventricular repolarization could be expected in ARVD on the basis of the anatomic alterations of the right ventricle and, perhaps, of regional abnormalities of sympathetic innervation, as observed in patients with ARVD.20 A complex multipolar QRST integral map has been related to local disparities of ventricular repolarization and thus to cardiac state of vulnerability to arrhythmias.6–10 In the present study, no patient showed a clear multipolar QRST map. A multipolar distribution probably reflects only gross regional disparities of the recovery process and may not represent a sufficiently sensitive index for minor heterogeneities.
In a previous study on patients with idiopathic long-QT syndrome, we were able to detect minor repolarization disparities by applying principal component analysis of the ST-T waveforms.21 Therefore, we decided to analyze the morphology of all the recorded ST-T waves using the same method in patients with ARVD. We assumed that a low value of the first or even of the first three components, corresponding to a relatively high information content of the remaining components, indicates a greater-than-normal variety of the ST-T waves, a likely expression of a more complex recovery process, caused by spatial and temporal disparities of repolarization.
In our study population, the values of the first component were significantly lower in patients with VT than in patients without VT (Table 2⇑). In particular, a low value (<69%) of component 1 resulted the best predictor of the occurrence of VT, being present in 6 of 9 VT patients, but only in 1 patient without VT. In fact, we estimated the predictive value of component 1 together with other variables (areas of negative values on QRST maps, negative T waves in leads V2 and V3, QRS duration in lead V1 >110 ms, QTc >420 ms, late potentials, and ε wave), some of which are commonly considered in clinical routine ECG, by applying a logistic regression analysis.14 A significant association with the occurrence of sustained VT was found only for component 1 (Table 3⇓).
Ventricular late potentials showed the same sensitivity (67%) as component 1 in identifying patients with VT, but they were present also in 54% of patients without VT. These findings are in agreement with data reported by other authors who demonstrated that late potentials are more closely related to the extent of right ventricular disease than with the occurrence of ventricular arrhythmias.22,23
In 8 patients, the duration of QRS complex was >110 ms in lead V1, a criterion proposed as diagnostic marker of ARVD.24 In these patients, the intraventricular conduction disturbances could have induced “secondary” alterations of repolarization. Nevertheless, principal component analysis of ST-T waves revealed an abnormally low value of component 1 in only 3 of 8 patients with QRS >110 ms in V1 (Table 1⇑). These data suggest that the heterogeneities of repolarization, as reflected by a low value of component 1, are probably poorly influenced by ventricular conduction disturbances.
The value of the first principal component was not correlated with age, sex, and QT duration in normal subjects, whereas a significant correlation (P<.05) was observed between component 1 and age in ARVD patients. This was probably due to the fact that the more diseased patients, who had a low component 1, were also older than the other patients. In ARVD patients, we did not find a significant correlation between a low value of component 1 and QT interval, presence of negative T waves, ε wave, and QRS duration in 12-lead ECG or ventricular late potentials in signal-averaged ECGs. In some cases (patients 4, 5, 9, and 22 in Table 1⇑), a low value of component 1 enabled us to detect abnormalities and possibly heterogeneities of repolarization not apparent in either the QRST maps or the 12-lead ECG.
This study has the following potential limitations
It involves a relatively small number of patients; therefore, the results need to be confirmed in a larger group.
Most patients suffering from VT were taking antiarrhythmic drugs at the time of body-surface map recording. In particular, amiodarone, which affects the repolarization process, could also influence the value of the principal components of ST-T waves. Unfortunately, we do not have body-surface map recordings in ARVD patients before and after treatment with amiodarone, and, to the best of our knowledge, no data have been reported on this subject. Specifically, we do not know the effects of amiodarone on the indexes of heterogeneities of repolarization as component 1. However, it has been reported that amiodarone prolongs the QT interval but does not increase or even reduces the repolarization dispersion, as can be measured from the differences between maximal and minimal QT or QTc values in the precordial or in the 12-lead ECG.25,26 In our laboratory, we had the opportunity of recording body-surface maps in 1 patient with left ventricular hypertrophy before and after acute amiodarone therapy (900 mg IV plus 2000 mg orally over 40 hours). In this patient, component 1 was similar before and after treatment (77% versus 74%, respectively). In 2 other patients with mild arterial hypertension on long-term treatment with amiodarone for preventing atrial fibrillation, the value of component 1 was within normal range. On the basis of all these data, we feel that amiodarone should not increase repolarization heterogeneities and that therefore it should not have reduced per se the value of component 1 in patients 1, 2, and 6 of our study population (Table 1⇑). On the other hand, in patient 3 who showed a normal value of component 1, it is impossible to know whether component 1 was already normal before treatment or was increased by amiodarone.
Concerning the 3 other patients treated with class 1c drugs who had normal component 1 values, some pharmacological effects on repolarization cannot be excluded, even though it is known that these effects should be mild. Thus, we recorded body-surface maps in 2 patients with paroxismal atrial fibrillation before and after propafenone; QRST maps and principal component analysis gave similar results before and after treatment.
A minority of patients (n=9) underwent electrophysiological study. Therefore, we did not perform a statistical comparison between the various indexes of vulnerability to arrhythmias deduced from repolarization analysis and the presence or absence of VT inducibility. Specifically, 6 patients in the VT group underwent electrophysiological study: in 5 patients, 3 with an abnormally low value of component 1 and 2 with a normal value, VT was inducible; in the remaining patient, who had an abnormally low component 1, VT was not induced. In 3 other patients without sustained VT and with normal values of component 1, VT was not inducible.
Principal component analysis provides a better quantitative assessment of the “complexity” of repolarization than other ECG measurements widely used in clinical practice. When applied to ARVD patients, principal component analysis of the ST-T waves recorded from the entire chest surface revealed abnormalities not detected by conventional ECG that can be considered indexes of arrhythmia vulnerability. If the present results are confirmed in a larger population of ARVD patients, principal component analysis of repolarization could contribute to the identification of patients at higher risk for life-threatening arrhythmias.
- Received June 25, 1997.
- Revision received September 3, 1997.
- Accepted September 23, 1997.
- Copyright © 1997 by American Heart Association
Han J, Moe GK. Nonuniform recovery of excitability in ventricular muscle. Circ Res. 1964;14:44–60.
Kuo CS, Munakata K, Reddy CP, Surawicz B. Characteristics and possible mechanism of ventricular arrhythmias dependent on the dispersion of action potential durations. Circulation. 1983;67:1356–1367.
Marcus FI, Fontaine GH, Guiraudon G, Frank R, Laurenceau JL, Malergue C, Grosgogeat Y. Right ventricular dysplasia: a report of 24 adult cases. Circulation. 1982;65:384–387.
Urie PM, Burgess MJ, Lux RL, Wyatt RF, Abildskov JA. The electrocardiographic recognition of cardiac states at high risk of ventricular arrhythmias. Circ Res. 1978;42:350–358.
Abildskov JA, Green LS, Lux RL. Detection of disparate ventricular repolarization by means of the body surface electrocardiogram. In: Zipes DP, Jalife J, eds. Cardiac Electrophysiology and Arrhythmias. Orlando, Fla: Grune and Stratton; 1985:495–499.
De Ambroggi L, Bertoni T, Locati E, Stramba-Badiale M, Schwartz PJ. Mapping of body surface potentials in patients with the idiopathic QT syndrome. Circulation. 1986;74:1334–1345.
Gardner MJ, Montague TJ, Armstrong CS, Horacek BM, Smith ER. Vulnerability to ventricular arrhythmias: assessment by mapping of body surface potentials. Circulation. 1986;73:684–692.
Mitchell LB, Hubley-Kozey CL, Smith ER, Wyse DG, Duff HJ, Gillis AM, Horacek BM. Electrocardiographic body surface mapping in patients with ventricular tachycardia: assessment of utility in the identification of effective pharmacological therapy. Circulation. 1992;86:383–393.
McKenna WJ, Thiene G, Nava A, Fontaliran F, Blomstrom-Lundqvist C, Fontaine G, Camerini F. Diagnosis of arrhythmogenic right ventricular dysplasia/cardiomyopathy. Br Heart J. 1994;71:215–218.
SippensGroenewegen A, Spekhorst H, van Hemel NM, Herre Kingma J, Hauer NW, Janse MJ, Dunning AJ. Body surface mapping of ectopic left and right ventricular activation: QRS spectrum in patients without structural heart disease. Circulation. 1990;82:879–896.
Horan LG, Flowers NC, Brody DA. Principal factor waveforms of the thoracic QRS complex. Circ Res. 1964;19:131–145.
Hosmer DW, Lemeshow S. Applied Logistic Regression. New York, NY: John Wiley & Sons: 1989.
Montague TJ, Smith ER, Cameron DA, Rautaharju PM, Klassen GA, Felmington CS, Horaceck BM. Isointegral analysis of body surface maps: surface distribution and temporal variability in normal sujects. Circulation. 1981;63:1166–1172.
Hirai M, Hayashi H, Ichihara Y, Adachi M, Kondo K, Suzuki A, Saito H. Body surface distribution of abnormally low QRST areas in patients with left ventricular hypertrophy: an index of repolarization abnormalities. Circulation. 1991;84:1505–1515.
Abildskov JA, Evans AK, Lux RL, Burgess MJ. Ventricular recovery properties and QRST deflection area in cardiac electrograms. Am J Physiol. 1980;239:H227–H231.
Abildskov JA. Effects of activation sequence on the local recovery of ventricular excitability in the dog. Circ Res. 1976;38:240–243.
Wichter T, Hindricks G, Lerch H, Bartenstein P, Borggrefe M, Shober O, Breithardt G. Regional myocardial sympathetic dysinnervation in arrhythmogenic right ventricular cardiomyopathy: an analysis using 123I-Meta-Iodobenzylguanidine scintigraphy. Circulation. 1994;89:667–683.
Hii JTY, Wyse DJ, Gillis AM, Duff HJ, Solylo MA, Mitchell LB. Precordial QT interval dispersion as a marker of torsade de pointes: disparate effects of class 1a antiarrhythmic drugs and amiodarone. Circulation. 1992;86:1376–1382.