To reduce QT measurement error, a new method was tested in which high-gain, high-speed, simultaneous 12-lead electrocardiographic (ECG) recordings were obtained during a single cardiac cycle. To increase its predictive power, the utility of combining QTD with the QRS duration for predicting susceptibility to ventricular tachyarrhythmia (VT) was analyzed. A total of 113 patients referred for electrophysiological study underwent baseline simultaneous 12-lead ECG followed by electrophysiological study to determine VT inducibility. Twenty-six patients had inducible VT while 87 patients did not. QT intervals and the width of QRS complex were measured from a single cardiac cycle with high-gain (8 times normal) and high-speed (100 mm/second) 12-lead ECG recordings.

T

his method resulted in 100% QT interval identification throughout all 12 leads for every patient. Receiver-operator characteristic curves (ROC) and the areas under the ROC curves (AUC) were used to quantitatively analyze the performance of four ECG variables (QTD3, QTD12, QTD12 + QRS and QTD3 + QRS).

All four ECG variables were significantly increased in the patients with inducible VT as compared to those without inducible VT. The QTD3 algorithm was less useful than QTD12 in predicting inducible VT; however, the addition of QRS duration to all QTD algorithms enhanced VT detection. In conclusion, we found: 1) QRS duration has an incremental benefit in the detection of VT when combined with QTD; 2) QTD12 + QRS duration provided the highest predictive power among the four tested algorithms; 3) high-gain, high-speed 12-lead ECG recordings reduced QT measurement error.

QT dispersion (QTD) has been proposed as an indicator of risk for ventricular tachyarrhythmias (VT) in different clinical settings (long QT syndrome,2,3 congestive heart failure4 and post-myocardial infarction5–9). However, a major limitation in the use of QTD is the intra- and inter-observer variability of the measurement,10–13 as well as its relatively low predictive power compared with other clinical indexes (late potentials, T-wave alternans and left ventricular ejection fraction).14 Previously, Cohen and colleagues demonstrated the utility of combining QTD with QRS duration in predicting susceptibility to VT.14 However, this technique was limited by the measurement of QT intervals from different cardiac cycles using a standard, 3-channel, 12-lead electrocardiogram (ECG). In addition, patients with bundle branch block were not excluded in the study and the contribution of QRS to the overall detection performance may have been obscured. The main objective of the present study was to examine the utility of combining QRS duration with QTD, measured from either 12-lead or 3-lead ECG, for predicting susceptibility to VT and to quantitatively compare their predictive performances.

METHODS

Table 1

Characteristics of patients with and without inducible sustained, monomorphic ventricular tachycardia.

One hundred and twenty-eight consecutive patients were referred for an electrophysiological study (EPS) to assess for VT. Fifteen patients with bundle branch block were excluded. The remaining 113 patients (65 men and 48 women; mean age ± SD, 66 ± 14 years), regardless of their cardiac substrates and initial presentations to electrophysiological study, were enrolled in this study as shown in Table 1.

All patients underwent a baseline simultaneous 12-lead ECG using a Prucka electrophysiology recording system (Prucka Inc., Houston, Texas) in the electrophysiology laboratory prior to programmed electrical stimulation off antiarrhythmic drugs. The Prucka system uses 12 channels to record a 12-lead ECG simultaneously; therefore, ECG variables can be measured in a single cardiac cycle. In addition, the recorded ECGs are stored in a digital format via computer hard disk and can be played back and displayed on a computer screen for review and analysis in a variety of gains and sweep speeds. All ECG variables were manually measured from the computer screen directly using an internal electronic caliper provided by the Prucka recording system. The electronic caliper is a software tool for measuring intervals on the computer screen directly without the need to print out paper strips. ECG measurements included QRS duration (QRS), QTD (QTmax-QTmin) in all 12 leads (QTD12) or in 3 quasi-orthogonal leads (leads I, aVF, V1) (QTD3). QTD3 + QRS and QTD12 + QRS were calculated based on QTD12, QTD3 and QRS measurements. All the ECG variables were measured in a single cardiac cycle with high gain (8 times normal) and high speed (100 mm/second) to facilitate more precise discrimination. QRS duration was measured from the lead with the earliest onset to the lead with the latest offset from all 12 leads. The end of the T-wave was measured where it intersected the isoelectric TP baseline. In cases of low T-wave amplitude, gain was further increased (16 times or 32 times normal) so that the end of the T-wave could be visually identified. In the presence of a U-wave, the end of the T-wave was obtained from the nadir between the T- and U-wave peaks. In the presence of biphasic T-waves, the intersection of the late stage of the T-wave either above or below the isoelectric TP baseline was used as the end of the T-wave. All measurements were performed by a single reader (WQ) who was blinded to the clinical characteristics of the patient.

Table 2

Summary of electrocardiographic measurements (mean ± SEM; all in ms) EPS = electrophysiology study; QT int Max = largest QT interval from 12-lead electrocardiogram; QT int Min = smallest QT interval from 12-lead electrocardiogram; QTD12c = dispersion in corrected QT interval from 12-lead electrocardiogram

Programmed electrical stimulation was performed using a standard protocol of up to three extra stimuli from two right ventricular sites (apex, outflow tract or septum using two different drive cycle lengths). Pacing stimuli were two milliseconds in duration and twice the diastolic threshold. VT susceptibility was defined as spontaneous and/or inducible sustained VT. Sustained VT was defined as VT of >= 30 seconds in duration or requiring cardioversion because of hemodynamic instability.

To measure the predictive power of the addition of QRS to QTD, the area under the curve (AUC) of the receiver operating characteristic (ROC) curve was computed for each of the four standard ECG algorithms as a predictor of VT and for QTD with QRS as combined predictors of VT. Receiver operator characteristic curves were determined. The areas below the ROC curves were simply calculated using the trapezoidal rule.18 To minimize the estimation error of the area, the maximum (Table 2) possible number of points were used to construct the ROC curves. A comparison of the areas of the four receiver operator characteristic curves was utilized in order to help define the optimal VT detection algorithm. All data are presented as means ± SEM unless otherwise noted and statistical comparisons of the categorical variables between inducible and non-inducible patients were done by the Chi-square analysis (with p < 0.05 considered statistically significant) or a Fisher exact test. A Satterthwaite adjustment was used when the variances of the two groups were unequal and differences of means are expressed as 95% confidence intervals.

The data were used to construct a multiple logistic regression model to predict the occurrence of VT based on QTD and QRS. The output of the model generated for each of the two explanatory variables, odds ratios as well as a p-value are an incremental contribution to the model. Due to the continuous nature of the independent variables, odds ratios are for increments of 30 (approximating the standard deviation).

RESULTS

Figure 1 (A)

(A) Receiver-operator characteristic curves of four standard electrocardiographic algorithms in full range and domain. The ROC of QRS is shown as a reference.

(B)

(B) Receiver-operator characteristic curves of four standard electrocardiographic algorithms in a specified range and domain for clarification at maximum output performance. The ROC of QRS is shown as a reference.

Sustained monomorphic VT was inducible in 26 of 113 patients (23%). The clinical characteristics, cardiac substrates and initial presentations of the patients with and without inducible sustained monomorphic VT are compared in Table 1. There were no significant differences in age, cardiac substrates [except coronary artery disease (CAD)], and initial presentations in inducibility of sustained VT. In these 113 patients, VT tended to be more likely induced in men than in women (19 of 46 men, 7 of 41 women; odds ratio, 2.4) and in patients with CAD compared to those without CAD (16 of 51 CAD patients,10 of 62 non-CAD patients; odds ratio, 1.9). However, these comparisons did not reach statistically significance.

Table 2 shows a summary of ECG measurements. Four patients had low T-wave amplitudes requiring an increase in gain to 16 or 32 to identify the end of the T-wave. Twenty-eight patients had a visible U-wave, and 10 had biphasic T-waves. All measurements listed except RR intervals are reasonable indexes for discrimination of VT inducibility. The QRS duration, mean QT interval, maximum QT interval, minimum QT interval and QT dispersion (both QTD3 and QTD12) were all significantly greater in patients with inducible VT compared to those without inducible VT. Specifically, QRS duration increased by 5.8%. QT interval increased by 8%, maximum QT interval increased by 10.4%, minimum QT interval increased by 6.3%, QTD3 increased by 48.6%, QTD12 increased by 35.7%, QTD12c increased by 36%, QTD3 + QRS increased by 23.2%, QTD12 + QRS increased by 20%, respectively, in patients with inducible VT as compared to those without inducible VT. Particularly, the mean minimum QT interval was increased in patients with inducible VT as compared to those with non-inducible studies (from 379 ± 4.7 ms to 403 ± 10.2 ms); the mean maximum QT interval also increased in patients with positive studies as compared with those with negative studies (from 441 ± 5.9 ms to 487 ± 12.3 ms). However, the increase in mean maximum QT interval (10.4%) was greater than the increase in the mean minimum QT interval (6.3%), which resulted in an increased mean QTD in patients with inducible VT. QTD12 (75.7 ± 9.7 ms) was significantly greater than QTD3 (49.3 ± 6.3 ms) in the entire patient population (p = 0.002). The logistic regression results indicate that both QTD (p = 0.005) and QRS (p = 0.04) make a statistically significant contribution to the model predicting VT. Both odds ratios differ significantly from 1. Thus, QRS added independent information to that contained in QTD.

Figure 2 (A)

(A) The area indexes measured as the area under the receiver-operator characteristic curves for full range and domain. QTD12 + QRS that has the highest area index is associated with the best algorithm for predicting susceptibility to VT due to the addition of QRS.

(B)

(B) The area indexes measured as the area under the receiver-operator characteristic curves for the specified range and domain. QTD12 + QRS still has the highest area index as shown due to the addition of QRS.

Table 3 shows a comparison of the four standard ECG algorithms at 70% specificity. The QTD12 method yields 16% higher sensitivity (58%) compared to the QTD3 method (50%). These results show that QTD12 + QRS and QTD3 + QRS algorithms had the highest sensitivity (73%) and highest positive and negative predictive values, while QTD3 had the lowest sensitivity (50%) and the lowest positive and negative predictive values. Specifically, by adding QRS to QTD12, sensitivity was improved from 58% to 73% (a 26% improvement). The positive and negative predictive values were also improved by 13.5% (from 0.37 to 0.42) and 2.4% (from 0.85 to 0.87), respectively. Similarly, by adding QRS to QTD3, sensitivity was improved from 50% to 73% (a 46% improvement). The positive and negative predictive values improved by 27% (from 0.33 to 0.42) and 6.1% (from 0.82 to 0.87), respectively. Thus, unlike the disparity in predictive values when only QTD3 and QTD12 were used, the addition of QRS duration to QTD demonstrated that both algorithms had similar output performances.

The ROC curve provides a graphic representation of the trade-off between false-positive rates and true-positive rates. It is a valuable tool for the assessment of the accuracy of diagnostic tests. Figure 1 shows ROC curves of QTD3, QTD12, QTD3 + QRS and QTD12 + QRS algorithms. As a comparison, QRS duration was also included in the figure. Figure 1A shows the full range of false detection rates. As a comparison, Figure 1B shows the ROC curves in the partial range of false detection rates from 0.2 to 0.5. Figure 2 demonstrates a comparison of the area indexes measured as area under the ROC curve (AUC). The AUC represents an overall performance of the ROC curve. In Figure 2A, the AUC was calculated based on Figure 1A (full range of false detection rate). The QTD12 method yields a 9% higher AUC (0.708) compared to the QTD3 method (0.649). AUC measured by the QTD3 + QRS method (0.715) is 10% higher than that measured by the QTD3 alone (0.649; Figure 2). Similarly, AUC measured by the QTD12 + QRS method (0.743) is 5% higher than that measured by the QTD12 alone (0.708; Figure 2) and is 4% higher than that measured by the QTD3 + QRS. Overall, the QTD12 + QRS had the highest area index of the four algorithms evaluated in this study. However, the ROC curves crossed at two points and thus the greater AUC for the combined QTD + QRS (0.74) was not significantly different than the AUC for QTD alone (0.71). In Figure 2B, the AUC was calculated based on Figure 1B (false detection rate from 20% to 50%). The QTD12 method yields a 41% higher AUC (0.196) compared to the QTD3 method (0.14). AUC measured by the QTD3 + QRS method (0.214) is 53% higher than that measured by the QTD3 alone (0.14). Similarly, AUC measured by the QTD12 + QRS method (0.23) is 17% higher than that measured by the QTD12 alone (0.196) and is 7% higher than that measured by the QTD3 + QRS (0.214). From Figures 2A and 2B, one can observe that adding QRS to QTD3 or QTD12 improves the detection performance of QTD3 or QTD12 if they are used alone, especially if detection is operated in the range of a 20–50% false detection rate. The difference is so obvious and would be statistically significant if we could legitimately test it. However, the p-value for the comparison is only meaningful if the false-positive rate range is chosen prior to rather than after inspection of the results. Therefore, we did not calculate the p-value. Figure 3A shows a linear regression analysis and correlation of QTD3 and QTD12 (R = 0.76), which shows a moderate correlation between QTD3 and QTD12. Figure 3B shows a linear regression analysis and correlation of QRS and QTD12 (R = 0.19), which shows no apparent correlation between QTD and QRS.

DISCUSSION

Figure 3 (A)

(A) A regression analysis performed with QTD12 as compared to QTD3 showing a moderately close correlation (R = 0.76).

(B)

(B) A regression analysis performed with QTD12 as compared to QRS showing no apparent correlation (R = 0.19).

Previously, Glancy and colleagues,16 as well as Cohen and colleagues,14 compared QTD measured from a small set of ECG leads (such as the three quasi-orthogonal leads of aVF, I and V1) with those measured from standard 12 leads. Their results demonstrated the consistency in discrimination of positive and negative outcomes using these two different measurements. However, it is not certain if a reduced set of ECG leads in QTD measurement could replace a standard 12-lead measurement. In the present study, we also demonstrated that QTD3 and QTD12 consistently predicted susceptibility to VT. However, we also demonstrated that QTD12 was a better discriminator of VT than QTD3 as evaluated by many criteria, such as sensitivity, AUC, positive and negative predictive values. The QTD12 method yielded a 16% higher sensitivity, a 9% higher AUC, a 12% higher positive predictive value and a 3.6% higher negative predictive value as compared to the QTD3 method (Table 3 and Figure 2). In the present study, we used a high-gain and high-speed digital ECG that resulted in 100% QT interval identification in all 12 leads of every ECG as compared to a mean of 9.9 leads as reported by Glancy and colleagues.16 This may be the major contributing factor that makes QTD12 superior to QTD3 in the present study. QTD reflects the regional discrepancies of ventricular recovery time. Therefore, the leads picking up maximum and minimum QT intervals may vary from patient to patient depending on their heart substrates. In the current study, the maximum QT intervals were heavily concentrated on V2, I and V6 (38.3%) while the minimum QT intervals were heavily concentrated on V1, aVL and V2 (53.2%). As a consequence, QTD measured from the quasi-orthogonal leads (I, aVF and V1) may miss a large percentage of maximum and minimum QT intervals. Besides the different measurement technique used in these studies, the discrepancy may also reflect different patient populations.

Table 3

A comparison of four standard electrocardiogram algorithms at 70% specifity.

Cohen and colleagues demonstrated14 that the QTD3 + QRS duration algorithm was as good as the QTD12 + QRS duration method for predicting VT. The present study also demonstrated that QRS duration enhanced the performance of QTD in predicting inducibility of VT. However, QTD12 + QRS duration performed better than QTD3 + QRS duration, as shown in their AUC measurements; AUC for QTD12 + QRS (0.708) was 5% greater than that for QTD3 + QRS (0.715). This is in line with the fact that QTD12 was superior to QTD3 if they were used alone. If the “partial area” of the AUC of the two ROC curves between the false-positive rate range of 0.20 and 0.60 in Figure 1 could be legitimately tested, the results would be significant as extensively investigated by McClish20 and Obuchowshki.21 Of course, the p-value for the comparison is only meaningful if the false-positive rate range is chosen prior to rather than after inspection of the ROCs. Thus, this p-value is not calculated, but only noted that the results were suggestive of increased clinical utility for the combination QTD + QRS as a combined predictor of VT when compared to QTD alone.

QRS and QTD reflect different electrophysiological mechanisms. However, it is not certain if QRS and QTD are independent indicators of susceptibility to VT. Brembilla-Perrot19 observed a strong correlation between inducibility of sustained VT and QRS duration. In the present study, we demonstrated that QRS duration may be a weak indicator of VT as compared to QTD. QTD12 yields 13.6% higher AUC compared to the QRS duration measurement. Ducceschi and co-workers15 observed a significant correlation between QTD and the presence of late potentials, as well as QTD and the mean QT interval. In the present study, we also observed a close correlation between QTD and mean QT interval (r = 0.52; p < 0.0001), as well as QRS duration and mean QT interval (r = 0.4; p < 0.0001). This result may suggest that global discrepancies of ventricular recovery time as represented by QTD, as well as general conduction velocity represented by QRS duration, are related to general repolarization duration represented by the mean QT interval.17 However, we did not observe a close relationship between QTD and QRS duration (r = 0.19; p = 0.04; Figure 3B). Our results also suggest that QRS duration and QTD are rather independent indicators of susceptibility to VT. This may be explained by the fact that QRS and QTD do not share the same electrophysiological mechanism; QRS duration is an index of conduction time in the ventricle while the QTD represents a global measurement of heterogeneity in ventricular repolarization.

In conclusion, QTD12 alone appears to be superior to QTD3 for discrimination of VT inducibility as supported by the logistic regression for the statistical utility of adding QRS to QTD. In addition, QTD12 + QRS provided the highest predictive power amongst all four tested algorithms for identifying VT. Also, the high-gain, high-speed, simultaneous 12-lead ECG recording appears to provide more reliable QTD measurements. However, it will be necessary to examine a second, independent sample to prospectively test the hypothesis that QRS contributes to increased clinical utility as well.