Abstract
Background: T‐wave morphology parameters based on principal component analysis (PCA) are candidate to better understand the relation between QT prolongation and torsades de pointes. We aimed to assess the repeatability and to determine the influence of T‐end position on PCA parameters.
Methods: Digital ECGs recorded from 30 subjects were used to assess short term (5 minutes), circadian and long‐term (28 days) repeatability of PCA parameters. The T‐end cursor position was moved backward and forward (± 8 ms) from its optimal position. We calculated QRS‐T angle, PCA ratio, and T‐wave residuum (TWR).
Results: At long‐term evaluation, coefficients of variation were 11.3 ± 9.9%, 11.7 ± 7.1%, and 23.0 ± 22.0% for the QRS‐T angle, PCA ratio, TWR, respectively. After moving the T‐end cursor, repeatability was 0.42 ± 0.2%, 1.00 ± 1.04%, 4.0 ± 4.2% for the same PCA parameters.
Conclusions: T‐wave morphology parameters based on PCA are reproducible with the exception of TWR and QRS‐T angle. In addition, PCA is robust, showing only little dependence on T‐end cursor position. These data should be taken into account for safety pharmacology trials.
Keywords: ECG, QT interval, principal component analysis, repeatability
The principal component analysis (PCA) is originally a technique that aims to represent a large number of signals by means of a limited number of fundamental values. When applied to digital ECG tracings, the method determines the “principal components” which represent most of the ECG information. Usually, the first three principal components provide nearly the total energy of the ECG. 1 , 2 Since the mathematical procedure calculates fundamental orthogonal components, PCA analysis of the ECG signal is a modern approach to the old‐fashioned vectocardiography based on orthogonal X, Y, and Z physical leads. 2 , 3
Clinical applications of PCA applied to digital ECGs are rapidly increasing. For instance, PCA analysis can compensate for respiration or body position‐related ECG morphology changes. 4 The PCA analysis can also enrich the characterization of ventricular repolarization. In particular, it has been shown that PCA analysis of body surface potential maps can identify non‐dipolar content. 1 , 5 , 6 Lux and De Ambroggi suggested that a complex multipeak distribution of ECG information might reflect regional ventricular repolarization heterogeneity. 1 , 5
Subsequently different authors have hypothesized that PCA analysis of the standard digital 12‐lead ECG may also reflect regional ventricular repolarization heterogeneities 7 , 8 , 9 thus, representing a better ECG marker of arrhythmic risk than the discredited so‐called “QT dispersion”. 9 , 10 , 11 Actually, different T‐wave parameters obtained after using PCA analysis, such as the QRS‐T angle, PCA ratio, and the T‐wave residuum have been shown as promising predictors of cardiovascular 12 , 13 , 14 or total mortality 13 , 15 , 16 in population studies.
Cardiovascular safety of new drug entities is critical. 17 , 18 , 19 Current guidelines on the proarrhythmic potential of a drug are based on QT prolongation, 20 but there is a poor understanding of the relation between QT prolongation and clinical risk of pro‐arrhythmia. We lack predictive biomarkers to improve our approach of the pathway leading to torsades de pointes. 18 , 21 PCA characterization of ventricular repolarization could represent a solution to replace QT duration.
However, before being used in pharmacological studies, any biomarker needs clinical qualification. The repeatability of T‐wave morphology parameters based on PCA over time as well as the influence of preprocessing require to be better documented. 22 T‐wave morphology parameters based on PCA are believed to be less dependent on precise determination of T wave offset than more traditional analyses, 9 , 23 although the influence of precise cursor position has not been systematically tested.
The aim of our study was two‐fold: (1) to assess the repeatability over time of T‐wave morphology parameters based on PCA in the setting of a thorough QT study, (2) to determine the influence of T‐end position on the same parameters.
PATIENTS AND METHODS
Study Design
The data reported in the present paper are part of a phase 1 clinical trial designed as a 3‐way crossover study. The study was randomized, double blind and placebo‐controlled. Only placebo data are reported in the present paper. The study was approved by two independent ethics committees and subjects signed an informed consent form.
The study population included 30 healthy subjects, 15 males, and 15 females. The mean age was 29 ± 5 years.
ECG Recordings and Preprocessing
Digital 12‐lead ECGs (MAC5000 GE Medical systems, Milwaukee, USA) were repeatedly recorded in resting supine position and in this trial the healthy subjects were in fed conditions. The electrodes were placed at the same place for each ECG recording. From the ECG data base we selected the following subset for each subject: (1) five ECGs at Day 1 (D1), including three ECGs (5 minutes apart) at 8:00 a.m. time point, a second ECG at 2:00 p.m., and a third one at 8:00 p.m. (2) at Day 14 (D14) and Day 28 (D28) one ECG recorded at 8:00 a.m. That subset was used to assess repeatability of PCA parameters (Fig. 1), the short term (three ECGs 5 minutes apart), the circadian (three ECGs 6 hours apart), and the long term ones, respectively (three ECGs 2 weeks apart).
Figure 1.
Study design.* The circadian repeatability was assessed at the placebo period.
First electrical cardiac signals from MAC5000 10‐second ECG strips were averaged (custom‐averaging process based on a sample‐to‐sample averaging). Then the position of QRS onset, QRS offset, and T‐wave end cursors was automatically determined on the averaged ECG waveforms and manually corrected by an expert cardiologist when necessary using the 12‐lead overlapped view the ECG (Fig. 2). PCA analysis was performed separately on the area of the QRS complex (QRS onset–QRS offset) and on the T‐wave time interval (QRS offset‐ T‐wave offset).
Figure 2.
ECG processing Left panel: Cursor position identification on the 12‐lead overlapped ECG Right panel: The three first PCA leads.
Finally, the position of the T‐wave end cursor was shifted from its “optimal” position as defined above on one of the ECG recorded at 8:00 a.m. The cursor position was moved 4 and 8 ms backward and forward. We thus obtained for each ECG five different cursor positions (QT‐8 ms, QT‐4 ms, QT, QT+4 ms, and QT+8 ms). The corresponding QT intervals were used to assess the influence of the T‐wave end position on PCA derived values.
PCA Analysis
PCA analysis was run using a custom dedicated software. Singular value decomposition was applied to the covariance matrix of the raw ECG data (amplitude at sampling time × ECG channel) from the eight independent ECG leads. Singular value decomposition was applied to obtain the eigenvectors and the eigenvalues of the QRS complex and of the T wave. 24 , 25 , 26
The orthogonal eigenvectors delineate a eight‐dimensional space, whereas the eigenvalues (λ1, ..., λ8) represent the energy in each dimension. The first dimension, given by the first eigenvector, contains the maximum energy. The sum of all the eigenvalues represents the total power of the ECG signal. Singular value decomposition was applied independently on QRS complex and T wave to obtain the eigenvectors and the eigenvalues for each part of the signal. The angle θ between QRS and T is defined as the angle between the principal vector of the QRS complex, and the principal vector extracted from the T wave.
The projection of the T wave on the subspace defined by the last five eigenvectors is commonly named the non‐dipolar components of the T wave or “T‐wave residuum” (TWR). The energy of the TWR is computed by the following formula: TWR4 − 8 = 100 −[∑i=4 8λi/∑i=1 8λi)* 100]. Similarly, the proportion of energy remaining after the first two orthogonal leads is calculated by TWR3 − 12 = 100 −[(∑i=38λi/∑i=18λi)*100].
In each dimension, the dispersion of the T wave PCA loop is given by sigma (σi, expressed in mV) where σi is defined as:
, i = 1…8 (Fig. 3).
Figure 3.
The T‐wave loop (upper panel) and its projection onto the PCA1/PCA2 and PCA1/PCA3 preferential planes.
The ratio of the second to the first σ (σ2/σ1) is the so‐called PCA ratio and provides an estimate of the roundness of the PCA T‐wave loop (Fig. 3).
The percentage of the PCA T‐wave loop projected on the plane defined by the two first eigenvectors is defined as: %T_σ1σ2 =
Similarly, 
represents the percentage of the PCA T‐wave loop projected on the space defined by the three first eigenvectors.
Statistical Analysis
Data are presented as mean ± SD. Quantitative data were compared using ANOVA for repeated measures. The coefficient of variation was defined as the intrasubject SD divided by the grand mean, multiplied by 100. In addition, Bland and Altman plots were calculated for the short‐term repeatability (10 minutes apart), long‐term repeatability (28 days), circadian repeatability (8 a.m. vs 2 p.m.). Bland and Altman method was also used to assess the influence of T‐wave end position.
RESULTS
Gender Differences
Table 1 shows the ECG and T‐wave morphology parameters based on PCA separately in males and females. Females had a non‐significant faster heart rate and prolonged QT/QTc intervals.
Table 1.
T‐Wave Morphology Parameters Based on PCA in Male and Female Normal Subjects
| Females | Males | |
|---|---|---|
| R‐R (ms) | 1042 ± 103 | 1091 ± 143 |
| [870; 1250] | [845; 1333] | |
| QT (ms) | 403 ± 25 | 399 ± 28 |
| [364; 446] | [358; 442] | |
| QTc Bazett (ms) | 395 ± 19 | 383 ± 23 |
| [349; 418] | [340; 412] | |
| QRS‐T angle | 51.9 ± 36 | 45.8 ± 18 |
| [16.9; 88.9] | [6.9; 67.7] | |
| %T_σ1σ2 | 92.6 ± 2.2 | 94.3 ± 1.5* |
| [88.9; 95.5] | [91.5; 96.4] | |
| %T_σ1σ2σ3 | 95.7 ± 1.4 | 96.7 ± 0.7* |
| [92.6; 97.4] | [95.3; 97.8] | |
| TWR 3–8 (%) | 0.260 ± 0.179 | 0.161 ± 0.135 |
| [0.060; 0.604] | [0.037; 0.489] | |
| TWR 4–8 (%) | 0.078 ± 0.060 | 0.042 ± 0.020* |
| [0.021; 0.209] | [0.019; 0.083] | |
| σ2/σ1 | 0.20 ± 0.07 | 0.23 ± 0.07 |
| [0.08; 0.31] | [0.12; 0.35] |
* P < 0.05
Mean ± SD – Numbers in brackets indicate the range
More than 90% of the T‐wave loop was included in two dimensions (%T_ σ1σ2) and more than 95% in three dimensions (%T_σ1σ2σ3). Both the %T_σ1σ2 and %T_σ1σ2σ3 were significantly higher in males than in females. Accordingly, the T‐wave residuum (TWR4–8) was significantly higher females.
The width of the T‐wave loop was close to 20% of its length (σ2/σ1) without significant difference between genders. The QRS‐T angle was around 45–50 degrees.
Repeatability of PCA Parameters over Time
Table 2 shows the actual values of ECG and T‐wave morphology parameters for the different time points. The values measured around 8 a.m. (either triplicate ECGs or ECGs across study days) were very close to each other. Oppositely, the variations of both ECG and T‐wave morphology parameters from 8 a.m. to 2 p.m. were significant.
Table 2.
Values of T‐Wave Morphology Parameters Based on PCA at Different Times
| Short‐Term Repeatability (n = 30) | ||||
|---|---|---|---|---|
| ECG 1 | ECG 2 | ECG 3 | ANOVA | |
| R‐R (ms) | 1065 ± 126 | 1042 ± 108 | 1048 ± 102 | NS |
| QT (ms) | 401 ± 23 | 399 ± 20 | 399 ± 22 | NS |
| QRS‐T angle | 49.7 ± 27.6 | 49.8 ± 26.8 | 49.2 ± 27.0 | NS |
| %T_σ1σ2 | 93.5 ± 2.2 | 92.6 ± 2.4 | 93.3 ± 2.6 | NS |
| %T_σ1σ2σ3 | 96.2 ± 1.2 | 96.3 ± 1.3 | 96.1 ± 1.5 | NS |
| TWR 3–8(%) | 0.207 ± 0.186 | 0.210 ± 0.213 | 0.228 ± 0.241 | NS |
| TWR 4–8(%) | 0.064 ± 0.054 | 0.065 ± 0.071 | 0.073 ± 0.090 | NS |
| σ2/σ1 | 0.21 ± 0.07 | 0.21 ± 0.07 | 0.21 ± 0.07 | NS |
| Long‐Term Repeatability (n = 29) | ||||
| D1 | D14 | D28 | ANOVA | |
| R‐R (ms) | 1064 ± 124 | 1057 ± 155 | 1068 ± 126 | NS |
| QT (ms) | 400 ± 23 | 400 ± 30 | 401 ± 26 | NS |
| QRS‐T angle | 50.3 ± 27.8 | 49.2 ± 27.9 | 48.2 ± 28.4 | NS |
| %T_σ1σ2 | 93.5 ± 2.3 | 93.2 ± 2.3 | 93.5 ± 2.1 | NS |
| %T_σ1σ2σ3 | 96.2 ± 1.2 | 96.2 ± 1.1 | 96.3 ± 1.2 | NS |
| TWR 3–8 (%) | 0.206 ± 0.189 | 0.241 ± 0.258 | 0.209 ± 0.166 | NS |
| TWR 4–8 (%) | 0.064 ± 0.065 | 0.059 ± 0.041 | 0.059 ± 0.048 | NS |
| σ2/σ1 | 0.21 ± 0.07 | 0.20 ± 0.08 | 0.21 ± 0.07 | NS |
| Circadian Repeatability (n = 30) | ||||
| 8 a.m. | 2 p.m. | 8 p.m. | ANOVA | |
| R‐R (ms) | 1043 ± 129 | 941 ± 129 | 1045 ± 150 | P < 0.001 |
| QT (ms) | 399 ± 29 | 388 ± 27 | 395 ± 28 | P < 0.001 |
| QRS‐T angle | 48.5 ± 27.2 | 52.4 ± 25.5 | 47.5 ± 27.2 | p < 0.001 |
| %T_σ1σ2 | 93.1 ± 2.5 | 91.8 ± 3.1 | 93.7 ± 2.0 | P < 0.001 |
| %T_σ1σ2σ3 | 96.1 ± 1.2 | 95.4 ± 1.5 | 96.4 ± 1.0 | P < 0.001 |
| TWR 3–8 (%) | 0.258 ± 0.316 | 0.415 ± 0.491 | 0.189 ± 0.157 | P < 0.001 |
| TWR 4–8 (%) | 0.064 ± 0.053 | 0.105 ± 0.102 | 0.049 ± 0.032 | P < 0.001 |
| σ2/σ1 | 0.21 ± 0.07 | 0.24 ± 0.11 | 0.19 ± 0.07 | P < 0.01 |
The co‐efficients of variation together with the Bland and Altman values are shown Table 3 and Figure 4. The repeatability of T‐wave morphology parameters slightly decreased from short to long‐term evaluation. The circadian repeatability was not as good as the long term one measured at the same time of the day. The TWR showed much higher coefficients of variability than other PCA parameters.
Table 3.
Repeatability over Time of ECG and T‐Wave Morphology Parameters
| Coefficient of Variation | Bland and Altman | |||||
|---|---|---|---|---|---|---|
| 10 minute | 4 weeks | Circadian | 10 minute | 4 weeks | Circadian | |
| QT | 1.08 ± 0.74 | 2.48 ± 1.75 | 2.51 ± 1.45 | 1.7 | −0.4 | 11.7 |
| % | % | % | (−11.5; 14.9) | (−26.8; 26.0) | (−17.0; 44.4) | |
| QTc | 1.58 ± 0.95 | 2.12 ± 1.44 | 2.44 ± 1.66 | −1.0 | 0.6 | −8.8 |
| % | % | % | (−19.5; 17.4) | (−20.8; 22.0) | (−32.9; 15.3) | |
| QRS‐T angle | 3.0 ± 2.1 | 11.3 ± 9.9 | 9.0 ± 6.5 | 0.5 | 2.2 | −3.9 |
| % | % | % | (−3.9; 4.8) | (−13.3; 17.6) | (−21; 13) | |
| %T_σ1σ2 | 0.4 ± 0.4 | 0.9 ± 0.6 | 1.3 ± 0.8 | 0.13 | 0.0 | 1.3 |
| % | % | % | (−1.49; 1.76) | (−2.1; 2.1) | (−2.2; 4.8) | |
| %T_σ1σ2σ3 | 0.3 ± 0.2 | 0.4 ± 0.2 | 0.7 ± 0.4 | 0.08 | −0.08 | 0.8 |
| % | % | % | (−0.89; 1.05) | (−1.15; 0.99) | (−0.9; 2.5) | |
| TWR 3–8 | 16.1 ± 18.6 | 36.4 ± 43.9 | 48.6 ± 69.2 | −0.02 | −0.001 | −0.16 |
| (%) | % | % | % | (−0.19; 0.15) | (−0.26; 0.25) | (−0.81; 0.49) |
| TWR 4–8 | 19.8 ± 30.3 | 23.0 ± 22.0 | 45.5 ± 55.8 | −0.009 | −0.004 | −0.04 |
| (%) | % | % | % | (−0.09; 0.07) | (−0.03; 0.04) | (−0.18; 0.09) |
| σ2/σ1 | 3.9 ± 2.7 | 11.7 ± 7.1 | 19.1 ± 19.9% | 0.003 | 0.004 | −0.03 |
| % | % | (−0.02; 0.02) | (−0.07; 0.08) | (0.21; 0.15) | ||
Bland & Altman : bias (± 1.96*SD) ‐ circadian = 8 a.m. versus 2 p.m.
Figure 4.
Bland and Altman plots of short and long‐term repeatability of %T_σ1σ2σ3.
The Role of T‐Wave End Cursor Position
Table 4 present the values of PCA parameters after shifting the position of the T‐offset cursor. Although statistically significant, the differences observed were rather small (Table 4) and of a lesser magnitude than the differences between triplicate ECG recordings. Correspondingly, the repeatability of the measures was excellent (Table 5 and Fig. 5).
Table 4.
Values of T‐Wave Morphology Parameters Based on PCA for Different T‐End Cursor Positions
| QT‐8 ms | QT‐4 ms | QT | QT +4 ms | QT +8 ms | |
|---|---|---|---|---|---|
| R‐R (ms) | 1055 ± 151 | 1055 ± 151 | 1055 ± 151 | 1055 ± 151 | 1055 ± 151 |
| QT (ms) | 393 ± 29 | 397 ± 29 | 401 ± 29 | 405 ± 29 | 409 ± 29 |
| QRS‐T angle | 48.5 ± 27.5 | 48.6 ± 27.5 | 48.7 ± 27.6 | 48.7 ± 27.6 | 48.8 ± 27.6 |
| %T_σ1σ2 | 93.3 ± 2.2 | 93.3 ± 2.2 | 93.2 ± 2.3 | 93.2 ± 2.3 | 93.1 ± 2.3 |
| %T_σ1σ2σ3 | 96.2 ± 1.03 | 96.2 ± 1.03 | 96.2 ± 1.04 | 96.1 ± 1.04 | 96.1 ± 1.04 |
| TWR 3–8 (%) | 0.234 ± 0.246 | 0.238 ± 0.251 | 0.241 ± 0.253 | 0.243 ± 0.255 | 0.245 ± 0.255 |
| TWR 4–8 (%) | 0.057 ± 0.039 | 0.058 ± 0.040 | 0.059 ± 0.040 | 0.061 ± 0.041 | 0.062 ± 0.041 |
| σ2/σ1 | 0.203 ± 0.08 | 0.203 ± 0.08 | 0.203 ± 0.08 | 0.203 ± 0.08 | 0.204 ± 0.08 |
ANOVA P < 0.001 for all PCA parameters except for σ2/σ1
Table 5.
Repeatability of T‐Wave Morphology Parameters Based on PCA According to QT Interval Measurement
| Coefficient of Variation | Bland & Altman | |||
|---|---|---|---|---|
| −4 ms versus +4 ms | −8 ms versus +8 ms | −4 ms versus +4 ms | −8 ms versus +8 ms | |
| QRS‐T angle | 0.21 ± 0.12% | 0.42 ± 0.24% | −0.19 (−0.46; 0.07) | −0.39 (−0.92; 0.14) |
| %T_σ1σ2 | 0.05 ± 0.04% | 0.10 ± 0.08% | 0.09 (−0.08; 0.26) | 0.18 (−0.15; 0.51) |
| %T_σ1σ2σ3 | 0.04 ± 0.03% | 0.07 ± 0.05% | 0.07 (−0.04;0.18) | 0.12 (−0.07; 0.32) |
| TWR 3–8 | 1.4 ± 1.7% | 3.0 ± 2.8% | −0.005 (−0.023; 0.013) | −0.011 (−0.05; 0.027) |
| TWR 4–8 | 2.2 ± 2.2% | 4.0 ± 4.2% | −0.0024 (−0.008; 0.003) | −0.0045 (−0.015; 0.006) |
| σ2/σ1 | 0.49 ± 0.52% | 1.00 ± 1.04% | 0.0005 (−0.005; 0.006) | 0.0009 (−0.010; 0.012) |
Bland & Altman : bias (± 1.96*SD)
Figure 5.
Bland and Altman plots for %T_σ1σ2σ3 after 16 ms change in QT duration.
Rate influences
With the exception of the QRS‐T angle, T‐wave morphology parameters based on PCA were correlated with the R‐R interval (Table 6). However, although significant, the correlations were weak (less of 10% of T‐wave morphology parameters changes could be explained by R‐R changes). In addition, the slope (alpha coefficient) of the relationships were low. For instance, a 100 ms RR interval change would induce a 0.3% absolute change in the %T_σ1σ2σ3 parameter.
Table 6.
Correlations with R‐R Interval
| Alpha Co‐efficient | P for Alpha Co‐efficient | Bêta Co‐Befficient | R2 | |
|---|---|---|---|---|
| QT | 0.119 | <0.0001 | 275 | 0.39 |
| QRS‐T angle | NS | NS | NS | NS |
| %T_σ1σ2 | 0.005 | <0.0001 | 87.7 | 0.09 |
| %T_σ1σ2σ3 | 0.003 | <0.0001 | 93.0 | 0.10 |
| TWR 3–8 | −0.001 | <0.05 | 0.610 | 0.03 |
| TWR 4–8 | −9.5 10−5 | <0.01 | 0.166 | 0.04 |
| σ2/σ1 | 2.0 10−4 | <0.0001 | 0.01 | 0.12 |
NS = not significant
DISCUSSION
Main Results
Most of T‐wave morphology parameters based on PCA, but not TWR and QRS‐T angle display stable results over time in healthy subjects. Another important result is that the position on the T‐offset cursor has only little influence on T‐wave morphology parameters.
Significance of T‐Wave Morphology Parameters Based on PCA
The QRS‐T angle is the easiest PCA parameters to be understood intuitively. In the early 1930s, Wilson and coworkers tried to separate between T‐wave variations related to changes in the depolarization wave (so‐called “secondary” T‐wave changes) and those related to the repolarization itself (or “primary” T‐waves changes). Wilson therefore introduced the concept of the “ventricular gradient” which is calculated by summing the QRS and the T‐wave areas from orthogonal XYZ leads. The calculation of the QRS‐T angle using PCA is mathematically related to the ventricular gradient. Accordingly, an increase in QRS‐T angle has been demonstrated as a predictor of bad outcome. 14 , 27
However, assuming that regional rather than global ventricular heterogeneities are implicated in arrhythmia mechanisms, 28 the QRS‐T angle is probably not the best surrogate for arrhythmic risk stratification.
The PCA ratio is an old vectocardiographic parameter that describes the projection of the T‐wave loop onto its preferential plane. 3 , 29 The length of a normal T‐wave loop is five times longer than its width, leading to a PCA ratio around 20%. Several data suggest that a malfunctioning ventricular repolarization process is associated with an increase in the PCA ratio (an increase of T‐wave loop roundness) 30 , 31 that in turn is associated with an increased mortality. 12 , 14 , 16 The association between an increased T‐wave loop roundness and increased mortality is not fully elucidated.
PCA analysis of body surface potential maps clearly demonstrates that after the first three components the remaining energy is associated with the presence of electrical multipeaks. 1 , 5 , 6 From 12‐lead ECG, the so‐called TWR could represent the non‐dipolar components of the T‐wave energy and could reflect regional ventricular repolarization heterogeneities, 7 , 8 although this link has yet not been fully demonstrated from the standard 12‐lead set. As in previous studies, we found that the TWR stands for only a small amount (i.e. less than 1%) of total T‐wave energy in normal subjects. 7 , 8 , 9 , 15
Since it represents energy, the TWR is not a descriptor of the shape of the T‐wave loop. The projection of the T‐wave loop onto its preferential space (σ1‐σ2‐σ3) is described by the three eigenvectors. It is well known that 95% of the T‐wave loop is projected on the σ1‐σ2‐σ3 space. 3 , 30
Gender and Heart Rate Influences
T‐wave morphology parameters based on PCA display relatively large standard deviation due to different modulating factors. Gender modulates T‐wave parameters with more complex repolarization feature in females. Indeed, in females a lesser percentage of T‐wave loop can be projected in a single plane or space, when compared to males. Females also show a trend to an increased TWR. These results are consistent with those published by Smetana et al. 7 In addition, gender‐related differences in ventricular repolarization properties are well documented both at the experimental and clinical levels. 32 , 33
More critically, PCA of the T‐wave morphology is also dependent on heart rate. 8 To take into account this physiological modulation, Smetana et al. suggest running PCA analysis after heart rate stratification (“binning”) from long‐term ECG series. 8 In the present study, the slopes of the relationships between heart rate and PCA parameters are small (≤6*10−3) although significant, explaining only 5–10% of the distribution of the parameters (Table 6). It is smaller than in the reports from Smetana et al. 8 the discrepancy being probably related to the range of heart rate considered. In the report of Smetana et al. long‐term ECG recordings had been obtained, thus providing larger RR ranges than those obtained when recording only short term 10‐second resting ECGs. In addition, Smetana et al. showed that T‐wave parameters mainly change for RR intervals shorter than 700 ms. 8 Of note, the heart rate influences on T‐wave morphology parameters based on PCA are much lesser than observed for the crude QT duration (>10% with R2= 39% in our study). Our data suggest that for RR intervals included between 900 and 1200 ms a correction for heart rate changes is not mandatory for T‐wave morphology parameters analysis.
T‐wave Morphology Parameters Based on PCA Repeatability over Time
In our study, the repeatability of PCA measures with the exception of TWR only slightly decreased from short to long‐term evaluations. However, the circadian repeatability was not as good as the short or the long term ones. Previous studies have demonstrated autonomic influences on T‐wave parameters. 34 , 35 In this study, the main changes were observed at 2 p.m., that is in postprandial conditions. Thus, the circadian differences in T‐wave morphology parameters based on PCA may not be related only to circadian heart rate changes. 36
The repeatability of the measure of the TWR was lesser than other PCA parameters. Batdorf et al. have already reported that the 1‐month reproducibility of TWR is not very high. 22 In their study, the reproducibility of TWR measurement could be enhanced by increasing the number of beats used for QRST averaging. 22 In our experience, the amplitude of the TWR was not correlated with the level of noise (data not shown). Hence, the reduction of TWR variability when increasing the number of beats used for QRST averaging 22 is probably not related to an increase of the signal to noise ratio. Besides, it is not known whether averaging the signal over a long period might mask the non‐dipolar component of the T wave. Therefore, further studies are needed to clarify the TWR variability.
The repeatability over time of the T‐wave morphology parameters other than the TWR and QRS‐T angle can be considered as close to the one of QT/QTc intervals (Table 3) and thus good enough to be used in clinical trials, provided that the time profile of ECG sample is considered.
Robustness
The vectocardiography theory suggests that the loop structure would be not very sensitive to the position of T‐wave end cursor. Indeed, the extremities of the T‐wave signal only represent a small proportion of samples used to determine the preferential planes. 37 In the present study, we systematically assessed the influence of T‐end position by shifting the cursor 4 and 8 ms backward and forward. Our results show that the T‐wave morphology parameters based on PCA changes following cursor modification were clinically insignificant. The difficulty of an accurate and reproducible T‐wave end assessment has been long recognized. 23 This limitation is even more critical in case of T‐wave morphology abnormalities. Therefore, the relative independence of T‐wave morphology parameters on terminal samples of the T‐wave would simplify the procedures involved in manual measurements of ECG time intervals. Our results give a rationale for the use of automatic T‐end assessment for PCA. 7 , 8 The limited influence of the last millisecond of the T‐wave signal is in accordance with experimental data. Previous studies have shown that the degree of repolarization heterogeneity is dependent on cells differential properties within the ventricular wall 38 but not on crude QT interval duration. 39
Study Limitations
The use of an average beat may have provided different results than the use of a single beat but the averaging process was preferred in order to improve the signal to noise ratio. It should also be acknowledged that it is difficult to compare stability of measurements that have not the same unit or the same scale. In addition, although T‐wave morphology parameters based on PCA have been proposed to characterize ventricular repolarization complexity their relevance to drug‐induced cardiac toxicity has not been so far demonstrated. The present study reports only placebo data of a phase 1 clinical trial, we could thus not investigate the drug‐induced changes in PCA parameters. However, the evaluation of drugs' impact on ventricular repolarization is based on placebo‐controlled studies. It is therefore important to have some insight on the intrinsic variability of the PCA parameters on baseline conditions.
CONCLUSION
The repeatability over time of the T‐wave morphology parameters other than the TWR and QRS‐T angle can be considered as close to the one of QT/QTc intervals.
The T‐wave morphology parameters based on PCA show small heart rate influences in resting conditions. In addition, T‐wave morphology parameters are robust, showing only little dependence on the accuracy of T‐end determination.
These data should be taken into account for safety pharmacology trials.
REFERENCES
- 1. Lux RL, Evans AK, Burgess MJ, et al Redundancy reduction for improved display and analysis of body surface potential maps. I. Spatial compression. Circ Res 1981;49:186–196. [DOI] [PubMed] [Google Scholar]
- 2. Macfarlane PW, Edenbrandt L. 12‐lead vectorcardiography in ischemic heart disease. J Electrocardiol 1992;24 (Suppl):188–193. [DOI] [PubMed] [Google Scholar]
- 3. Rubel P, Fayn J, Mohsen N, et al New methods of quantitative assessment of the extent and significance of serial ECG changes of the repolarization phase. J Electrocardiol 1988;21(Suppl):S177–S181. [DOI] [PubMed] [Google Scholar]
- 4. Leanderson S, Laguna P, Sornmo L. Estimation of the respiratory frequency using spatial information in the VCG. Med Eng Phys 2003;25:501–507. [DOI] [PubMed] [Google Scholar]
- 5. De Ambroggi L, Bertoni T, Locati E, et al. Mapping of body surface potentials in patients with the idiopathic long QT syndrome. Circulation 1986;74:1334–1345. [DOI] [PubMed] [Google Scholar]
- 6. Montague TJ, Smith ER, Cameron DA, et al Isointegral analysis of body surface maps: Surface distribution and temporal variability in normal subjects. Circulation 1981;63:1166–1172. [DOI] [PubMed] [Google Scholar]
- 7. Smetana P, Batchvarov VN, Hnatkova K, et al Sex differences in repolarization homogeneity and its circadian pattern. Am J Physiol Heart Circ Physiol 2002;282:H1889–H1897. [DOI] [PubMed] [Google Scholar]
- 8. Smetana P, Batchvarov VN, Hnatkova K, et al Ventricular gradient and nondipolar repolarization components increase at higher heart rate. Am J Physiol Heart Circ Physiol 2004;286:H131–H136. [DOI] [PubMed] [Google Scholar]
- 9. Malik M, Acar B, Gang Y, et al QT dispersion does not represent electrocardiographic interlead heterogeneity of ventricular repolarization. J Cardiovasc Electrophysiol 2000;11:835–843. [DOI] [PubMed] [Google Scholar]
- 10. Kors JA, Van Herpen G, Van Bemmel JH. QT dispersion as an attribute of T‐loop morphology. Circulation 1999;99:1458–1463. [DOI] [PubMed] [Google Scholar]
- 11. Coumel P, Maison‐Blanche P, Badilini F. Dispersion of ventricular repolarization: Reality? Illusion? Significance? Circulation 1998;97:2491–2493. [DOI] [PubMed] [Google Scholar]
- 12. Okin PM, Devereux RB, Fabsitz RR, et al Principal component analysis of the T wave and prediction of cardiovascular mortality in American Indians: The strong heart study. Circulation 2002;105:714. [DOI] [PubMed] [Google Scholar]
- 13. Zabel M, Malik M, Hnatkova K, et al Analysis of T‐wave morphology from the 12‐lead electrocardiogram for prediction of long‐term prognosis in male US veterans. Circulation 2002;105:1066. [DOI] [PubMed] [Google Scholar]
- 14. Rautaharju PM, Kooperberg C, Larson JC. Electrocardiographic predictors of incident congestive heart failure and all‐cause mortality in postmenopausal women: The women's health initiative. Circulation 2006;113:481–9. [DOI] [PubMed] [Google Scholar]
- 15. Okin PM, Malik M, Hnatkova K, et al Repolarization abnormality for prediction of all‐cause and cardiovascular mortality in American Indians: The Strong Heart study. J Cardiovasc Electrophysiol 2005;16:945–951. [DOI] [PubMed] [Google Scholar]
- 16. Okin PM, Devereux RB, Lee ET, et al Strong heart study. Electrocardiographic repolarization complexity and abnormality predict all‐cause and cardiovascular mortality in diabetes: The strong heart study. Diabetes 2004;53:434–40. [DOI] [PubMed] [Google Scholar]
- 17. Crouch MA, Limon L, Cassano AT. Clinical relevance and management of drug‐related QT interval prolongation. Pharmacotherapy 2003;23:881–908. [DOI] [PubMed] [Google Scholar]
- 18. Vos MA, Van Opstal JM, Leunissen JD, et al Electrophysiologic parameters and predisposing factors in the generation of drug‐induced Torsade de Pointes arrhythmias. Pharmacol Ther 2001;92:109–122. [DOI] [PubMed] [Google Scholar]
- 19. Malik M, Camm AJ. Evaluation of drug‐induced QT interval prolongation: Implications for drug approval and labelling. Drug Saf 2001;24:323–351. [DOI] [PubMed] [Google Scholar]
- 20. International Conference on Harmonisation (ICH): The Clinical Evaluation of QT/QTc Interval Prolongation and Proarrhythmic Potential for Non‐antiarrhythmic Drugs ICH E14, Step 4, May 12, 2005. http://www.emea.eu.int/pdfs/human/ich/000204en.pdf [PubMed]
- 21. Redfern WS, Carlsson L, Davis AS, et al Relationships between preclinical cardiac electrophysiology, clinical QT interval prolongation and torsade de pointes for a broad range of drugs: Evidence for a provisional safety margin in drug development. Cardiovasc Res 2003;58:32–45. [DOI] [PubMed] [Google Scholar]
- 22. Batdorf BH, Feiveson AH, Schlegel TT. The effect of signal averaging on the reproducibility and reliability of measures of T‐wave morphology. J Electrocardiol 2006;39:266–270. [DOI] [PubMed] [Google Scholar]
- 23. Xue Q, Reddy S. Algorithms for computerized QT analysis. J Electrocardiol 1998;30(Suppl):181–186. [DOI] [PubMed] [Google Scholar]
- 24. Damen AA, Van Der Kam J. The use of the singular value decomposition in electrocardiography. Med Biol Eng Comput 1982;20:473–482. [DOI] [PubMed] [Google Scholar]
- 25. Acar B, Koymen H. SVD‐based on‐line exercise ECG signal orthogonalization. IEEE Trans Biomed Eng 1999;46:311–321. [DOI] [PubMed] [Google Scholar]
- 26. Acar B, Yi G, Hnatkova K, et al Spatial, temporal and wavefront direction characteristics of 12‐lead T‐wave morphology. Med Biol Eng Comput 1999;37:574–584. [DOI] [PubMed] [Google Scholar]
- 27. Kors JA, Rijnbeek P, Van Herpen G, Keulers PP, et al Spatial repolarization parameters for predicting cardiac death in the elderly. J Electrocardiol. 2004;37:198–200. [DOI] [PubMed] [Google Scholar]
- 28. Kuo CS, Munakata K, Reddy CP, et al Characteristics and possible mechanism of ventricular arrhythmia dependent on the dispersion of action potential durations. Circulation 1983;67:1356–1367. [DOI] [PubMed] [Google Scholar]
- 29. Fayn J, Rubel P. CAVIAR: A serial ECG processing system for the comparative analysis of VCGs and their interpretation with auto‐reference to the patient. J Electrocardiol 1988;21(Suppl):S173—S176. [DOI] [PubMed] [Google Scholar]
- 30. Badilini F, Fayn J, Maison‐Blanche P, et al Quantitative aspects of ventricular repolarization: Relationship between three‐dimensional T wave loop morphology and scalar QT dispersion. Ann Noninvasive Electrocardiol 1997;2:146–157. [Google Scholar]
- 31. Yi G, Prasad K, Elliott P, et al T wave complexity in patients with hypertrophic cardiomyopathy. Pacing Clin Electrophysiol 1998;21:2382–2386. [DOI] [PubMed] [Google Scholar]
- 32. Drici MD, Burklow TR, Haridasse V, et al Sex hormones prolong the QT interval and downregulate potassium channel expression in the rabbit heart. Circulation 1996;94:1471–4. [DOI] [PubMed] [Google Scholar]
- 33. Macfarlane PW, McLaughlin SC, Devine B, et al Effects of age, sex, and race on ECG interval measurements. J Electrocardiol 1995;27(Suppl):14–19. [DOI] [PubMed] [Google Scholar]
- 34. Batchvarov V, Dilaveris P, Farbom P, et al New descriptors of homogeneity of the propagation of ventricular repolarization. Pacing Clin Electrophysiol 2000;23:1968–1972. [DOI] [PubMed] [Google Scholar]
- 35. Ghuran A, Batchvarov V, Dilaveris P, et al Reflex autonomic modulation of automatically measured repolarization parameters. Pacing Clin Electrophysiol 2000;23:1973–1976. [DOI] [PubMed] [Google Scholar]
- 36. Dota CD, Edvardsson N, Schutzer KM, et al Inter‐ and intraday variability in major electrocardiogram intervals and amplitudes in healthy men and women. Pacing Clin Electrophysiol 2003;26:361–366. [DOI] [PubMed] [Google Scholar]
- 37. Rubel P, Fayn J, Mohsen N. Stability of surface T wave and corrected QT interval in normal male population In: Butros GS, Schwartz PJ. Clinical aspects of ventricular repolarization. Farrand Press; London 1989, pp. 58–66. [Google Scholar]
- 38. Antzelevitch C, Fish J. Electrical heterogeneity within the ventricular wall. Basic Res Cardiol 2001;96:517–527. [DOI] [PubMed] [Google Scholar]
- 39. Antzelevitch C, Oliva A. Amplification of spatial dispersion of repolarization underlies sudden cardiac death associated with catecholaminergic polymorphic VT, long QT, short QT and Brugada syndromes. J Intern Med 2006;259:48–58. [DOI] [PMC free article] [PubMed] [Google Scholar]