- Email: [email protected]
Should the JT rather than the QT interval be used to detect prolongation of ventricular repolarization?
Journal of Electrocardiology
Vol. 25 Supplement
Should the JT Rather Than the QT Interval Be Used to Detect Prolongation of Ventricular Repolarization? An Assessment in Normal Conduction and in Ventricular Conduction Defects
Sophia H. Zhou, PhD, Shirley Wong, MScN, Pentti M. Rautaharju, Nachiket Kamik, BSc, and Harry P. Calhoun, PhD
MD, PhD,
Abstract: It has been suggested that the JT rather than QT interval properly reflects repolarization duration in ventricular conduction defects (VCD). The authors examined the influence of QRS duration on the JT and QT intervals in 20,687 normal adult subjects and 2,865 subjects with various categories of VCD. Estimates for coefficients for multiple regression of QRS duration on QT and JT intervals combined with a correction term for heart rate (HR) were determined for each VCD category. QRS duration accounted for about 16% of total QT variation, but had a practically negligible effect on JT interval in complete bundle branch blocks. A single-parameter formula was derived for the JT prolongation index of the form JTI = JT(HR + 100)/5 18, with a JTI 2 112 identifying repolarization prolongation in all VCD categories. It is concluded that it is preferable to predict JT rather than QT as a more appropriate index of duration of repolarization in VCD. Key words: QRS duration, QT interval, JT interval, ventricular conduction defects.
It has been established that QRS duration has a highly significant, although in practical terms, a relatively modest influence on the QT interval in normal subjects with normal ventricular conduction. ’ This observation is of considerable theoretical interest in relation to the ventricular repolarization process be-
cause it indicates that the commonly advocated concept of reverse sequence of repolarization cannot be valid for all myocardial regions. Das evaluated the relationships of QRS duration and QT and JT intervals in 18 patients with right bundle branch block (RBBB) and in 24 patients with left bundle branch block (LBBB) in comparison with 33 normal subjects, and concluded that the prolongation of the rate-corrected QT interval in complete bundle branch blocks was entirely secondary to a prolonged depolarization time.2 Das concluded that the JT rather than QT interval is an appropriate measure of ventricular repolarization time. In view of the practical as well as theoretical im-
From the Cardiac Epidemiology Coordinating and Research Centre (EPICORE), Division of Cardiology, University of AlbeTta. Edmonton, Alberta, Canada.
Supported in part by the Alberta Heart and the Alberta Heritage Foundation for Reprint requests: Pentti M. Rautahaju, Medical Research Building, University of berta, Canada T6G 2S2.
and Stroke Foundation Medical Research. MD, PhD, 2 14 Heritage Alberta, Edmonton, Al-
131
132
Journal of Electrocardiology
Vol. 25 Supplement
portance of these issues, we examined the influence of QRS duration on QT and JT intervals in a large electrocardiographic data bank composed of various categories of complete and incomplete bundle branch blocks. Our primary goal was to derive prediction formulas properly reflecting ventricular repolarization time with appropriate correction for heart rate variations, while at the same time eliminating the bias produced by the variations in ventricular excitation time.
Materials
The group with various categories of ventricular conduction defects (VCD) was stratified into five subgroups of complete and incomplete bundle branch blocks using the Minnesota Code criteria. Complete bundle branch blocks include Code 7.1 (left bundle branch block or LBBB), Code 7.2 (right bundle branch block or RBBB), and 7.4 (IVCD, complete block of unspecified type), all require QRS duration 2 120 ms. Incomplete LBBB (Code 7.6) requires QRS duration of 1OO- 120 ms and incomplete RBBB (Code 7.3) requires QRS duration of 1OO- 120 ms with R’ > R pattern in lead V1 or Vz.
and Methods Electrocardiographic Measurements
Study Population
The source data for this study were derived from a large, epidemiologic electrocardiographic data pool of the EPICORE Centre at the University of Alberta. This data bank was composed of electrocardiographic records from community-based health surveys, including the second and third National Health and Nutrition Surveys (NHANES2 and 3) and the Hispanic Health and Nutrition Examination Survey (HHANES) . The composition of the study population is shown in Table 1. The group with normal ventricular conduction consisted of 20,687 adult subjects, aged 20 to 99 years, from these community-based populations. Excluded from the normal group were individuals with primary electrocardiographic abnormalities known to influence the QT interval, such as ventricular and atrioventricular conduction defects and various Minnesota Code3 items related to myocardial infarction and repolarization abnormalities (Minnesota Code 1.1-1.3, 5.1-5.3, and 4.1-4.3). Table 1. Composition of the Study Population in Various Categories of Ventricular Conduction Defects Classified According to the Minnesota Code Criteria Minnesota Code 7.0 7.1 7.2 7.4 7.3 7.6
Category Normal conduction Left bundle branch block Right bundle branch block Complete bundle branch block unspecified type (IVCD) Incomplete right bundle branch block Incomplete left bundle branch block
No. Subjects
Mean Age (years)
20,687 231 673
51.4 63.9 63.8
616
60.1
607
53.5
738
54.6
QT and JT measurements were made by the NOVACODE Computer Program4,5 from eight independent simultaneously sampled components of the standard 12-lead electrocardiograms. In one of the studies (NHANES2), the conventional electrocardiograms were sampled in groups of three simultaneous leads, and to avoid differences in QT measurements from different lead groups,6 QT and JT measurements in this subgroup were made from Frank orthogonal X, Y, and Z leads. A comparison of QT measurements in a large subset of eight simultaneously sampled components of the 12-lead electrocardiograms and the Frank leads showed no significant differences between these two sampling and measurement procedures. QT and JT measurements were made after selective averaging of normally conducted complexes contained in a lo-second electrocardiographic record. Proper detection of correct QRS onset and offset (J point) and T offset was visually verified using a highresolution graphics terminal. Leads with excessive noise were rejected and occasional wave detection errors with difficult waveforms were corrected by trained operators.
Statistical Methods
A statistical software program (SAS) was used for data analysis. Several linear multiple regression models were introduced to assess the contribution of QRS duration to QT and JT intervals in various study subgroups. In the first step, optimal correction was made for heart rate by introducing the so called rate factor (RF) into the QT interval, predicted from the formula, QT = 656/(1 + 0.01 HR). In a previous study this was observed to produce the best correc-
QT and JT in Ventricular
Conduction
Defects
Zhou et al.
l
133
tion for heart rate (HR) variations among 12 functional formulas evaluated. ’ The two primary regression equations used were: QT = ollRF + PIQRS (Model 1)
-0.3-
and
-0.4-
JT = cu2RF + P2QRS (Model 2)
Bz
where QT, JT, and QRS intervals are expressed in ms. Two additional regression models were subsequently used to evaluate JT prediction accuracy without the inclusion of QRS duration as a covariate:
-05
-o.& 0.2
JT = CX~ RF + y3 (Model 3)
I
0.3
0.4
I
0.5
I
0.6
1
I 0.7
0.8
I
0.9
fh
and the same formula without intercept: JT = cx4RF (Model 4). R-square values were determined to evaluate what fraction of the total QT or JT variance can be accounted for by each prediction formula, and rootmean-square errors were determined to assess the overall prediction accuracy.
Results Coefficient estimates for regression of QRS on QT and JT in multiple regression models together with the correction term for heart rate are listed in Table 2. For QRS regression on QT, the values of regression coefficients p1 (Model 1) range from 0.29 in incomplete LBBB to 0.99 for IVCD. The rate correction factor and QRS duration combined explain from 50% to 66% of the total QT variance, as seen from the R-square values for Model 1. QRS duration alone accounts for approximately 16% of the total QTirariante in complete bundle branch blocks, as seen from incremental R-square values with the inclusion of
Fig. 1. Coefficients of QRS duration regression on the QT (PI) and JT intervals (p2) from the multiple regression model containing the heart rate correction term (models 1 and 2) for various categories of conduction defects as defined in Table 1.
QRS duration in the multiple regression model. The R-square increment for normal conduction and for incomplete RBBB is considerably less than for complete bundle branch blocks-O.03 and 0.04, respectively-but still highly significant (p < 0.001 for both). On the other hand, QRS duration has no significant effect on QT in incomplete LBBB (p > 0.05). The incremental R-square values with the inclusion of QRS duration in the multiple regression model for JT in Table 2 reveal that QRS duration influence on JT is practically negligible in complete bundle branch blocks (p > 0.05). QRS duration retains its significant association with JT in normal conduction and in incomplete RBBB (p < 0.001 for both), and although still significant, in practical terms, is nearly negligible for incomplete LBBB (AR* = 0.01). Figure 1 summarizes the relationship between
Table 2. Coefficient Estimates for Regression of QRS on QT (& in Model 1) and on JT (Pz in Model 2) After Correction for Heart Rate in Normal Ventricular Conduction and in Five Categories of Ventricular Conduction Defects ElectrocardiographicCategory
a1
Pl
Y1
R=
AR2
~2
P2
Yz
R2
AR2
Normal Left bundle branch block Right bundle branch block Complete block of unspecified type Incomplete right bundle branch block Incomplete left bundle branch block
596 577 503 563 591
0.55 0.87 0.77 0.99 0.56
-7 -20 20 -38 -7
0.62 0.60 0.54 0.54 0.66
0.03 0.16 0.16 0.15 0.04
596 577 503 563 591
-0.45 -0.13 -0.23 -0.01 - 0.44
-6 -20 20 -38 -7
0.62 0.50 0.43 0.48 0.62
0.07 0 0.01 0 0.03
523
0.29
65
0.50
0
523
-0.71
65
0.50
0.0 1
The rate correction was performed by determining the regression coefficients 01 and az for the rate correction factor (RF), with RF = I/( 1 + 0.01 HR). R2 = squared multiple correlation coefficient. AR* = the increment in R2 values with the inclusion of QRS duration in QT and JT prediction formulas. QRS, QT, and JT intervals are in ms. Model 1: QT = a,RF + P,QRS + -y,. Model 2: JT = a2RF + PzQRS + YZ.
134
Journal of Electrocardiology
Vol. 25 Supplement
Table 3. Comparison of JT Prediction Accuracy of Single-parameter Formula (Model 4), Same Formula With Intercept (Model 3), Three-parameter QT, and JT Prediction Formulas (Models 1 and 2 from Table 2) Model 3 QT = opRF + y3
ECG Category Normal LBBB RBBB IVCD Incomplete RBBB Incomplete LBBB
a3
Y3
564 571 499 563 568 519
Model 4 JT = ofiF
Model 1 QT = oiRF + BrQRS + yr
Model 2 JT = azRF + PzQRS + yz
RMSE
a4
RMSE
RMSE
RMSE
-30 -36 -7 -39 -35
20 25 24 28 19
518 511 485 498 509
19 25 24 28 19
19 25 24 28 18
19 25 24 28 18
-8
20
507
20
20
20
RMSE (ms) = the root-mean-square error of prediction; RF = the rate correction factor defined in Table 2; LBBB = left bundle branch block; RBBB = right bundle branch block; IVCD = complete block of unspecified type.
regression coefficients p1 and p2 for QRS regression on QT and JT. It is noted that as the values of the QRS regression coefficient for QT progressively increase from 0.29 for incomplete LBBB to near unity for IVCD, the values for the QRS regression coeffcient for JT increase from - 0.71 for incomplete LBBB to near zero for IVCD in such a fashion that for all categories pZ = p1 - 1. This result is due to the functional relationship between the statistical models chosen for the analysis as will be pointed out in the discussion section. In comparing the performance of various statistical models for QT and JT prediction, R-square values do not reveal all the information needed for comparative evaluation. It is also important to consider prediction accuracy. Furthermore, it is desirable to search for models that yield an adequate prediction accuracy with the smallest number of parameters. Root-mean-square errors for all four statistical models used are listed in Table 3. It is seen that the single-parameter model for JT prediction (Model 4), containing only the correction factor for heart rate, performs as well as the other more complex models, for instance, the three-parameter model used for QT prediction (Model 1). The root-mean-square error for JT with this model is 19 ms for normal conduction, and ranges from 19 to 28 ms for various VCD categories. In view of the similarity of the JT prediction formulas for Model 4 in Table 3 in all VCD categories with the possible exception of RBBB, it was decided to evaluate what fraction of subjects in each category will be classified as prolonged JT according to the JT prediction formula for the normal group. This was decided by deriving the following formula for the JT prolongation index (JTI) : JTI = (JT/JT,)
X 100 = JT(HR + 100)/518,
where JT is the measured JT in ms and JT, = a4RF =
5 18/( 1 + O.OlHR) is the predicted JT for the normal group from Table 3. The upper 2.5 percentile normal limit for JTI turned out to be 112, that is, 12% above the mean JTI for the normal group. Table 4 indicates that JTI values in incomplete bundle branch blocks and in RBBB are within the range of normal JTI. The prevalence of JT prolongation was 4.3% and 3.6% in LBBB and IVCD. The question as to what extent these subgroups, with an apparent prolongation of repolarization time, represent primary repolarization abnormalities warrants a separate investigation.
Discussion The results from the present investigation reveal that QRS duration has a pronounced influence on the QT interval in complete bundle branch blocks, explaining about 16% of the total QT variance in these VCD categories. Like in normal conduction, QRS duration influence on QT interval is considerably weaker in incomplete RBBB, and not significant in incomplete LBBB. This investigation was designed to seek an answer to the question whether the JT rather than QT interTable 4. ST Prolongation Prevalence Category LBBB RBBB IVCD Incomplete RBBB Incomplete LBBB
No. Subjects
n
%
231 673 616 607 738
10 12 22 8 19
4.3 1.8 3.6 1.3 2.5
Prevalence of prolonged repolarization times in various categories of ventricular conduction defects defined by the JT prolongation index (JTI) that exceed the upper 2.5 percentile normal limit (JTI 2 112) according to the formula JTI = JT (HR + 100)/518 where JT is measured in ms. LBBB = left bundle branch block; RBBB = right bundle branch block; IVCD = complete block of unspecified type.
QT and JT in Ventricular
val should be used to detect alterations in repolarization time in VCD. Our results indicate that from a statistical point of view it is immaterial whether QT or JT is used in these prediction formulas as long as an adjustment is made for QRS duration in QT prediction. Inherent in this reasoning is the assertion that QT duration is linearly dependent on the ventricular excitation time, which also implies that the JT interval is linearly and inversely dependent on QRS duration. This fact explains the observed linear, inverse relationship between the coefficients of QRS regression on QT and JT (f31 and p2 in Fig. 1) since a linear term was used for QRS duration as a covariate and JT = QT - QRS. QRS duration is not associated with JT interval significantly in complete bundle branch blocks and, although statistically significant, the association is weak in incomplete bundle branch blocks. From a practical point of view, the formulas with the smallest number of parameters are more desirable than the more complex multi-parameter formulas as long as an adequate prediction accuracy can be maintained. Our results indicate (Table 3) that JT prediction can be achieved with a single-parameter formula (without QRS duration) with equal accuracy in terms of root-mean-square error as QT prediction with two or three-parameter formulas in all VCD categories. It has been suggested that the JT rather than QT interval is a proper measure of the duration of repolarization, particularly in VCD. To answer this question, it is necessary to clarify what is meant by repolarization duration in the global sense when estimated from body surface electrocardiograms. An equally relevant question is what is meant by dispersion of repolarization times. At a cellular level, two phases can be distinguished in the repolarization process; the slow phase (plateau) and the fast phase of repolarization (phase 3). In body surface electrocardiograms, the time demarcation point between the slow and the fast phase of repolarization is denoted by the often ill-defined onset of the T wave. Three different situations can be conceptually formulated to clarify these questions about the repolarization process in conduction defects and in normal conduction. 1. Repolarization sequence is truly and universally reverse with respect to excitation sequence. The basic requirements for this condition are that QRS duration has no inlluence on the QT interval and that the T wave is fully concordant with QRS. The condition, which comes closest to this situation, is the incomplete LBBB where the QRS regression coefficient on QT in Model 1 was small and insignificant.
Conduction
Defects
l
Zhou et al.
135
In incomplete LBBB, the increased thickness of the left ventricular wall may explain the increased QRS duration. If the reverse sequence truly prevails, then the JT interval reflects closest action potential durations at the endocardial regions of the left ventricle. The T wave duration in turn, reflects dispersion of the fast phase of repolarization times in the left ventricle. 2. Repolarization sequence is mainly reverse, but concordant sequence prevails during the terminal part of repolarization in myocardial regions depolarized last. This situation may come closest to reality in normal conduction and in incomplete right bundle branch block. While there is ample evidence that the repolarization sequence is reverse in the free wall of the left ventricle, the net repolarization process is likely to be concordant with the spatial and temporal sequence of excitation in myocardial regions excited during the terminal phase of QRS. In these conditions, the main portion of the QRS and T waves are concordant but the terminal portions may be discordant. Also, QRS duration will influence the QT interval since the excitation time of myocardial regions depolarized latest will influence the onset of repolarization in these regions that repolarize latest. Thus, the JT interval reflects action potential durations in the basal portions of the ventricles excited last. 3. Repolarization sequence is mainly concordant with depolarization sequence. In this situation, QRS duration has a strong influence on the QT interval and T waves are fully disconcordant with QRS. This appears to be the case in IVCD and to a considerable extent in LBBB and RBBB, provided that primary repolarization abnormalities are absent. Thus, in complete bundle branch blocks the JT interval reflects closest action potential durations in basal portions of the left or right ventricle or the intraventricular septum, depending on the type of excitation pattern. The QT interval in concordant sequence of repolarization reflects the sum of excitation time and the duration of repolarization in the basal portions of the ventricles. In conclusion, it would seem that the meaning of the QT and JT intervals, repolarization duration, and dispersion of repolarization times varies depending on the type of ventricular conduction. In general, the QT interval has no distinct electrophysiological meaning at the cellular level. Neither the QT nor the JT interval can be used as a measure of duration of repolarization if this is meant to reflect the interval between the earliest onset and latest offset of the whole global repolarization process. Dispersion of ventricular repolarization times, associated with the fast phase of repolarization is, probably best estimated by global T wave duration. In normal conduc-
136
Journal of Electrocardiology
Vol. 25 Supplement
tion, in complete bundle branch blocks, and in incompIete right bundle branch block, the JT interval may reflect closest action potential durations in the basal portions of the heart.
3.
4.
References 5. 1. Rautahaju PM, Warren JW, Calhoun HP: Estimation of QT prolongation: a persistent avoidable error in computer electrocardiography. J Electrocardiol 23 (suppl) : Ill,1991 2. Das G: QT interval and repolarization time in patients
6.
with intraventricular conduction delay. J Electrocardiol 23:49, 1990 Blackbum H, Keys A, Simonson E et al: The electrocardiogram in population studies: a classification system. Circulation 2 1: 1160, 1960 Rautahaju PM, MacInnis PJ, Warren JW et al: Methodology of ECG interpretation in the Dalhousie Program. NOVACODE ECG classification procedures for clinical trials and population health surveys. Methods Inf Med 29:362, 1990 Rautahaju PM, Calhoun HP, Chaitman BR: NOVACODE serial ECG classification system for clinical trials and epidemiological studies. J Electrocardiol 24: 179, 1992 Cowan JC, Yusoff K, Moore M et al: Importance of lead selection in QT interval measurement. Am J Cardiol6 1: 83, 1988
Vol. 25 Supplement
Should the JT Rather Than the QT Interval Be Used to Detect Prolongation of Ventricular Repolarization? An Assessment in Normal Conduction and in Ventricular Conduction Defects
Sophia H. Zhou, PhD, Shirley Wong, MScN, Pentti M. Rautaharju, Nachiket Kamik, BSc, and Harry P. Calhoun, PhD
MD, PhD,
Abstract: It has been suggested that the JT rather than QT interval properly reflects repolarization duration in ventricular conduction defects (VCD). The authors examined the influence of QRS duration on the JT and QT intervals in 20,687 normal adult subjects and 2,865 subjects with various categories of VCD. Estimates for coefficients for multiple regression of QRS duration on QT and JT intervals combined with a correction term for heart rate (HR) were determined for each VCD category. QRS duration accounted for about 16% of total QT variation, but had a practically negligible effect on JT interval in complete bundle branch blocks. A single-parameter formula was derived for the JT prolongation index of the form JTI = JT(HR + 100)/5 18, with a JTI 2 112 identifying repolarization prolongation in all VCD categories. It is concluded that it is preferable to predict JT rather than QT as a more appropriate index of duration of repolarization in VCD. Key words: QRS duration, QT interval, JT interval, ventricular conduction defects.
It has been established that QRS duration has a highly significant, although in practical terms, a relatively modest influence on the QT interval in normal subjects with normal ventricular conduction. ’ This observation is of considerable theoretical interest in relation to the ventricular repolarization process be-
cause it indicates that the commonly advocated concept of reverse sequence of repolarization cannot be valid for all myocardial regions. Das evaluated the relationships of QRS duration and QT and JT intervals in 18 patients with right bundle branch block (RBBB) and in 24 patients with left bundle branch block (LBBB) in comparison with 33 normal subjects, and concluded that the prolongation of the rate-corrected QT interval in complete bundle branch blocks was entirely secondary to a prolonged depolarization time.2 Das concluded that the JT rather than QT interval is an appropriate measure of ventricular repolarization time. In view of the practical as well as theoretical im-
From the Cardiac Epidemiology Coordinating and Research Centre (EPICORE), Division of Cardiology, University of AlbeTta. Edmonton, Alberta, Canada.
Supported in part by the Alberta Heart and the Alberta Heritage Foundation for Reprint requests: Pentti M. Rautahaju, Medical Research Building, University of berta, Canada T6G 2S2.
and Stroke Foundation Medical Research. MD, PhD, 2 14 Heritage Alberta, Edmonton, Al-
131
132
Journal of Electrocardiology
Vol. 25 Supplement
portance of these issues, we examined the influence of QRS duration on QT and JT intervals in a large electrocardiographic data bank composed of various categories of complete and incomplete bundle branch blocks. Our primary goal was to derive prediction formulas properly reflecting ventricular repolarization time with appropriate correction for heart rate variations, while at the same time eliminating the bias produced by the variations in ventricular excitation time.
Materials
The group with various categories of ventricular conduction defects (VCD) was stratified into five subgroups of complete and incomplete bundle branch blocks using the Minnesota Code criteria. Complete bundle branch blocks include Code 7.1 (left bundle branch block or LBBB), Code 7.2 (right bundle branch block or RBBB), and 7.4 (IVCD, complete block of unspecified type), all require QRS duration 2 120 ms. Incomplete LBBB (Code 7.6) requires QRS duration of 1OO- 120 ms and incomplete RBBB (Code 7.3) requires QRS duration of 1OO- 120 ms with R’ > R pattern in lead V1 or Vz.
and Methods Electrocardiographic Measurements
Study Population
The source data for this study were derived from a large, epidemiologic electrocardiographic data pool of the EPICORE Centre at the University of Alberta. This data bank was composed of electrocardiographic records from community-based health surveys, including the second and third National Health and Nutrition Surveys (NHANES2 and 3) and the Hispanic Health and Nutrition Examination Survey (HHANES) . The composition of the study population is shown in Table 1. The group with normal ventricular conduction consisted of 20,687 adult subjects, aged 20 to 99 years, from these community-based populations. Excluded from the normal group were individuals with primary electrocardiographic abnormalities known to influence the QT interval, such as ventricular and atrioventricular conduction defects and various Minnesota Code3 items related to myocardial infarction and repolarization abnormalities (Minnesota Code 1.1-1.3, 5.1-5.3, and 4.1-4.3). Table 1. Composition of the Study Population in Various Categories of Ventricular Conduction Defects Classified According to the Minnesota Code Criteria Minnesota Code 7.0 7.1 7.2 7.4 7.3 7.6
Category Normal conduction Left bundle branch block Right bundle branch block Complete bundle branch block unspecified type (IVCD) Incomplete right bundle branch block Incomplete left bundle branch block
No. Subjects
Mean Age (years)
20,687 231 673
51.4 63.9 63.8
616
60.1
607
53.5
738
54.6
QT and JT measurements were made by the NOVACODE Computer Program4,5 from eight independent simultaneously sampled components of the standard 12-lead electrocardiograms. In one of the studies (NHANES2), the conventional electrocardiograms were sampled in groups of three simultaneous leads, and to avoid differences in QT measurements from different lead groups,6 QT and JT measurements in this subgroup were made from Frank orthogonal X, Y, and Z leads. A comparison of QT measurements in a large subset of eight simultaneously sampled components of the 12-lead electrocardiograms and the Frank leads showed no significant differences between these two sampling and measurement procedures. QT and JT measurements were made after selective averaging of normally conducted complexes contained in a lo-second electrocardiographic record. Proper detection of correct QRS onset and offset (J point) and T offset was visually verified using a highresolution graphics terminal. Leads with excessive noise were rejected and occasional wave detection errors with difficult waveforms were corrected by trained operators.
Statistical Methods
A statistical software program (SAS) was used for data analysis. Several linear multiple regression models were introduced to assess the contribution of QRS duration to QT and JT intervals in various study subgroups. In the first step, optimal correction was made for heart rate by introducing the so called rate factor (RF) into the QT interval, predicted from the formula, QT = 656/(1 + 0.01 HR). In a previous study this was observed to produce the best correc-
QT and JT in Ventricular
Conduction
Defects
Zhou et al.
l
133
tion for heart rate (HR) variations among 12 functional formulas evaluated. ’ The two primary regression equations used were: QT = ollRF + PIQRS (Model 1)
-0.3-
and
-0.4-
JT = cu2RF + P2QRS (Model 2)
Bz
where QT, JT, and QRS intervals are expressed in ms. Two additional regression models were subsequently used to evaluate JT prediction accuracy without the inclusion of QRS duration as a covariate:
-05
-o.& 0.2
JT = CX~ RF + y3 (Model 3)
I
0.3
0.4
I
0.5
I
0.6
1
I 0.7
0.8
I
0.9
fh
and the same formula without intercept: JT = cx4RF (Model 4). R-square values were determined to evaluate what fraction of the total QT or JT variance can be accounted for by each prediction formula, and rootmean-square errors were determined to assess the overall prediction accuracy.
Results Coefficient estimates for regression of QRS on QT and JT in multiple regression models together with the correction term for heart rate are listed in Table 2. For QRS regression on QT, the values of regression coefficients p1 (Model 1) range from 0.29 in incomplete LBBB to 0.99 for IVCD. The rate correction factor and QRS duration combined explain from 50% to 66% of the total QT variance, as seen from the R-square values for Model 1. QRS duration alone accounts for approximately 16% of the total QTirariante in complete bundle branch blocks, as seen from incremental R-square values with the inclusion of
Fig. 1. Coefficients of QRS duration regression on the QT (PI) and JT intervals (p2) from the multiple regression model containing the heart rate correction term (models 1 and 2) for various categories of conduction defects as defined in Table 1.
QRS duration in the multiple regression model. The R-square increment for normal conduction and for incomplete RBBB is considerably less than for complete bundle branch blocks-O.03 and 0.04, respectively-but still highly significant (p < 0.001 for both). On the other hand, QRS duration has no significant effect on QT in incomplete LBBB (p > 0.05). The incremental R-square values with the inclusion of QRS duration in the multiple regression model for JT in Table 2 reveal that QRS duration influence on JT is practically negligible in complete bundle branch blocks (p > 0.05). QRS duration retains its significant association with JT in normal conduction and in incomplete RBBB (p < 0.001 for both), and although still significant, in practical terms, is nearly negligible for incomplete LBBB (AR* = 0.01). Figure 1 summarizes the relationship between
Table 2. Coefficient Estimates for Regression of QRS on QT (& in Model 1) and on JT (Pz in Model 2) After Correction for Heart Rate in Normal Ventricular Conduction and in Five Categories of Ventricular Conduction Defects ElectrocardiographicCategory
a1
Pl
Y1
R=
AR2
~2
P2
Yz
R2
AR2
Normal Left bundle branch block Right bundle branch block Complete block of unspecified type Incomplete right bundle branch block Incomplete left bundle branch block
596 577 503 563 591
0.55 0.87 0.77 0.99 0.56
-7 -20 20 -38 -7
0.62 0.60 0.54 0.54 0.66
0.03 0.16 0.16 0.15 0.04
596 577 503 563 591
-0.45 -0.13 -0.23 -0.01 - 0.44
-6 -20 20 -38 -7
0.62 0.50 0.43 0.48 0.62
0.07 0 0.01 0 0.03
523
0.29
65
0.50
0
523
-0.71
65
0.50
0.0 1
The rate correction was performed by determining the regression coefficients 01 and az for the rate correction factor (RF), with RF = I/( 1 + 0.01 HR). R2 = squared multiple correlation coefficient. AR* = the increment in R2 values with the inclusion of QRS duration in QT and JT prediction formulas. QRS, QT, and JT intervals are in ms. Model 1: QT = a,RF + P,QRS + -y,. Model 2: JT = a2RF + PzQRS + YZ.
134
Journal of Electrocardiology
Vol. 25 Supplement
Table 3. Comparison of JT Prediction Accuracy of Single-parameter Formula (Model 4), Same Formula With Intercept (Model 3), Three-parameter QT, and JT Prediction Formulas (Models 1 and 2 from Table 2) Model 3 QT = opRF + y3
ECG Category Normal LBBB RBBB IVCD Incomplete RBBB Incomplete LBBB
a3
Y3
564 571 499 563 568 519
Model 4 JT = ofiF
Model 1 QT = oiRF + BrQRS + yr
Model 2 JT = azRF + PzQRS + yz
RMSE
a4
RMSE
RMSE
RMSE
-30 -36 -7 -39 -35
20 25 24 28 19
518 511 485 498 509
19 25 24 28 19
19 25 24 28 18
19 25 24 28 18
-8
20
507
20
20
20
RMSE (ms) = the root-mean-square error of prediction; RF = the rate correction factor defined in Table 2; LBBB = left bundle branch block; RBBB = right bundle branch block; IVCD = complete block of unspecified type.
regression coefficients p1 and p2 for QRS regression on QT and JT. It is noted that as the values of the QRS regression coefficient for QT progressively increase from 0.29 for incomplete LBBB to near unity for IVCD, the values for the QRS regression coeffcient for JT increase from - 0.71 for incomplete LBBB to near zero for IVCD in such a fashion that for all categories pZ = p1 - 1. This result is due to the functional relationship between the statistical models chosen for the analysis as will be pointed out in the discussion section. In comparing the performance of various statistical models for QT and JT prediction, R-square values do not reveal all the information needed for comparative evaluation. It is also important to consider prediction accuracy. Furthermore, it is desirable to search for models that yield an adequate prediction accuracy with the smallest number of parameters. Root-mean-square errors for all four statistical models used are listed in Table 3. It is seen that the single-parameter model for JT prediction (Model 4), containing only the correction factor for heart rate, performs as well as the other more complex models, for instance, the three-parameter model used for QT prediction (Model 1). The root-mean-square error for JT with this model is 19 ms for normal conduction, and ranges from 19 to 28 ms for various VCD categories. In view of the similarity of the JT prediction formulas for Model 4 in Table 3 in all VCD categories with the possible exception of RBBB, it was decided to evaluate what fraction of subjects in each category will be classified as prolonged JT according to the JT prediction formula for the normal group. This was decided by deriving the following formula for the JT prolongation index (JTI) : JTI = (JT/JT,)
X 100 = JT(HR + 100)/518,
where JT is the measured JT in ms and JT, = a4RF =
5 18/( 1 + O.OlHR) is the predicted JT for the normal group from Table 3. The upper 2.5 percentile normal limit for JTI turned out to be 112, that is, 12% above the mean JTI for the normal group. Table 4 indicates that JTI values in incomplete bundle branch blocks and in RBBB are within the range of normal JTI. The prevalence of JT prolongation was 4.3% and 3.6% in LBBB and IVCD. The question as to what extent these subgroups, with an apparent prolongation of repolarization time, represent primary repolarization abnormalities warrants a separate investigation.
Discussion The results from the present investigation reveal that QRS duration has a pronounced influence on the QT interval in complete bundle branch blocks, explaining about 16% of the total QT variance in these VCD categories. Like in normal conduction, QRS duration influence on QT interval is considerably weaker in incomplete RBBB, and not significant in incomplete LBBB. This investigation was designed to seek an answer to the question whether the JT rather than QT interTable 4. ST Prolongation Prevalence Category LBBB RBBB IVCD Incomplete RBBB Incomplete LBBB
No. Subjects
n
%
231 673 616 607 738
10 12 22 8 19
4.3 1.8 3.6 1.3 2.5
Prevalence of prolonged repolarization times in various categories of ventricular conduction defects defined by the JT prolongation index (JTI) that exceed the upper 2.5 percentile normal limit (JTI 2 112) according to the formula JTI = JT (HR + 100)/518 where JT is measured in ms. LBBB = left bundle branch block; RBBB = right bundle branch block; IVCD = complete block of unspecified type.
QT and JT in Ventricular
val should be used to detect alterations in repolarization time in VCD. Our results indicate that from a statistical point of view it is immaterial whether QT or JT is used in these prediction formulas as long as an adjustment is made for QRS duration in QT prediction. Inherent in this reasoning is the assertion that QT duration is linearly dependent on the ventricular excitation time, which also implies that the JT interval is linearly and inversely dependent on QRS duration. This fact explains the observed linear, inverse relationship between the coefficients of QRS regression on QT and JT (f31 and p2 in Fig. 1) since a linear term was used for QRS duration as a covariate and JT = QT - QRS. QRS duration is not associated with JT interval significantly in complete bundle branch blocks and, although statistically significant, the association is weak in incomplete bundle branch blocks. From a practical point of view, the formulas with the smallest number of parameters are more desirable than the more complex multi-parameter formulas as long as an adequate prediction accuracy can be maintained. Our results indicate (Table 3) that JT prediction can be achieved with a single-parameter formula (without QRS duration) with equal accuracy in terms of root-mean-square error as QT prediction with two or three-parameter formulas in all VCD categories. It has been suggested that the JT rather than QT interval is a proper measure of the duration of repolarization, particularly in VCD. To answer this question, it is necessary to clarify what is meant by repolarization duration in the global sense when estimated from body surface electrocardiograms. An equally relevant question is what is meant by dispersion of repolarization times. At a cellular level, two phases can be distinguished in the repolarization process; the slow phase (plateau) and the fast phase of repolarization (phase 3). In body surface electrocardiograms, the time demarcation point between the slow and the fast phase of repolarization is denoted by the often ill-defined onset of the T wave. Three different situations can be conceptually formulated to clarify these questions about the repolarization process in conduction defects and in normal conduction. 1. Repolarization sequence is truly and universally reverse with respect to excitation sequence. The basic requirements for this condition are that QRS duration has no inlluence on the QT interval and that the T wave is fully concordant with QRS. The condition, which comes closest to this situation, is the incomplete LBBB where the QRS regression coefficient on QT in Model 1 was small and insignificant.
Conduction
Defects
l
Zhou et al.
135
In incomplete LBBB, the increased thickness of the left ventricular wall may explain the increased QRS duration. If the reverse sequence truly prevails, then the JT interval reflects closest action potential durations at the endocardial regions of the left ventricle. The T wave duration in turn, reflects dispersion of the fast phase of repolarization times in the left ventricle. 2. Repolarization sequence is mainly reverse, but concordant sequence prevails during the terminal part of repolarization in myocardial regions depolarized last. This situation may come closest to reality in normal conduction and in incomplete right bundle branch block. While there is ample evidence that the repolarization sequence is reverse in the free wall of the left ventricle, the net repolarization process is likely to be concordant with the spatial and temporal sequence of excitation in myocardial regions excited during the terminal phase of QRS. In these conditions, the main portion of the QRS and T waves are concordant but the terminal portions may be discordant. Also, QRS duration will influence the QT interval since the excitation time of myocardial regions depolarized latest will influence the onset of repolarization in these regions that repolarize latest. Thus, the JT interval reflects action potential durations in the basal portions of the ventricles excited last. 3. Repolarization sequence is mainly concordant with depolarization sequence. In this situation, QRS duration has a strong influence on the QT interval and T waves are fully disconcordant with QRS. This appears to be the case in IVCD and to a considerable extent in LBBB and RBBB, provided that primary repolarization abnormalities are absent. Thus, in complete bundle branch blocks the JT interval reflects closest action potential durations in basal portions of the left or right ventricle or the intraventricular septum, depending on the type of excitation pattern. The QT interval in concordant sequence of repolarization reflects the sum of excitation time and the duration of repolarization in the basal portions of the ventricles. In conclusion, it would seem that the meaning of the QT and JT intervals, repolarization duration, and dispersion of repolarization times varies depending on the type of ventricular conduction. In general, the QT interval has no distinct electrophysiological meaning at the cellular level. Neither the QT nor the JT interval can be used as a measure of duration of repolarization if this is meant to reflect the interval between the earliest onset and latest offset of the whole global repolarization process. Dispersion of ventricular repolarization times, associated with the fast phase of repolarization is, probably best estimated by global T wave duration. In normal conduc-
136
Journal of Electrocardiology
Vol. 25 Supplement
tion, in complete bundle branch blocks, and in incompIete right bundle branch block, the JT interval may reflect closest action potential durations in the basal portions of the heart.
3.
4.
References 5. 1. Rautahaju PM, Warren JW, Calhoun HP: Estimation of QT prolongation: a persistent avoidable error in computer electrocardiography. J Electrocardiol 23 (suppl) : Ill,1991 2. Das G: QT interval and repolarization time in patients
6.
with intraventricular conduction delay. J Electrocardiol 23:49, 1990 Blackbum H, Keys A, Simonson E et al: The electrocardiogram in population studies: a classification system. Circulation 2 1: 1160, 1960 Rautahaju PM, MacInnis PJ, Warren JW et al: Methodology of ECG interpretation in the Dalhousie Program. NOVACODE ECG classification procedures for clinical trials and population health surveys. Methods Inf Med 29:362, 1990 Rautahaju PM, Calhoun HP, Chaitman BR: NOVACODE serial ECG classification system for clinical trials and epidemiological studies. J Electrocardiol 24: 179, 1992 Cowan JC, Yusoff K, Moore M et al: Importance of lead selection in QT interval measurement. Am J Cardiol6 1: 83, 1988