Author + information
- Received January 12, 1999
- Revision received April 21, 2000
- Accepted June 26, 2000
- Published online November 1, 2000.
- Markus Zabel, MD∗,* (, )
- Michael R Franz, MD, PhD, FACC‡,
- Thomas Klingenheben, MD†,
- Boris Mansion, MD†,
- Heinz-Peter Schultheiss, MD∗ and
- Stefan H Hohnloser, MD, FACC†
- ↵*Reprint requests and correspondence: Dr. Markus Zabel, Department of Medicine, Division of Cardiology, Klinikum Benjamin Franklin, Free University of Berlin, Hindenburgdamm 30, 12200 Berlin, Germany
The study was done to determine whether variables of QT dispersion from the 12-lead electrocardiogram (ECG) are dependent on heart rate.
The dispersion of the QT interval is under evaluation as a risk marker in patients at risk for ventricular arrhythmias. Assuming that a similar rate correction is necessary as for the QT interval itself, investigators have frequently reported QTc-dispersion values utilizing the Bazett formula. It is not known whether there is a physiologic basis for such a rate correction in the human heart.
In 35 patients referred for evaluation of ventricular arrhythmias, digital 12-lead ECGs recorded at various heart rates during submaximal exercise testing and again during atrial pacing upon electrophysiologic testing were submitted to computerized interactive analysis of several ECG dispersion variables.
Data from 11 patients were excluded due to incomplete high-quality analysis possible at all heart rates. From the remaining 24 patients, a total of 193 ECG recordings at various heart rates (ranging from 76 ± 17 beats/min to 117 ± 14 beats/min during atrial pacing and from 78 ± 18 beats/min to 110 ± 14 beats/min during exercise testing) were available. A highly significant linear relationship with heart rate was found for both the QT interval and the Q-to-T-peak interval. By contrast, standard QT interval dispersion (QTmax − QTmin), the T-peak-to-T-end interval, and the average area under the T wave did not change with increasing heart rates.
Dispersion of the QT interval and other ECG variables of dispersion of ventricular repolarization are independent of heart rate. Therefore, it is not necessary to rate-correct these measurements.
Dispersion of the QT interval is being evaluated as an easily accessible risk marker from the surface electrocardiogram (ECG) in patients prone to ventricular tachyarrhythmias (1–5). It has been shown to be useful in patients with the congenital (2) or the acquired (3) long QT syndrome. There have also been positive reports on its usefulness in patients with hypertrophic obstructive cardiomyopathy (6), in patients with congestive heart failure (4), and in patients after myocardial infarction (5,7), although the overall value of QT dispersion (QTD) is currently doubtful (8–11). The skepticism in weighing the body of literature available on QTD has been primarily about methodological concerns (8,9,11). In particular, the nonstandardized and inaccurate recording techniques as well as the insufficient analysis methodology have been criticized (8,9,11). Moreover, the often-used practice of rate correction of QTD has been seriously questioned (12). This custom was derived from the well-defined rate-dependence of the QT interval (13–17), which had been extrapolated to dispersion variables such as QTD and JT dispersion (2–7). Usually, rate-corrected variables have been calculated using the Bazett formula (13) and labeled QTc or JTc dispersion. A physiologic basis for this practice, however, has never been described. In an experimental model, we could previously demonstrate that the dispersion of ventricular repolarization (DVR) is independent of heart rate (HR) (18) and subsequently designed the present study in patients.
Thirty-five patients referred for evaluation of ventricular tachyarrhythmias and electrophysiologic testing were studied. A need for revascularization was ruled out by means of coronary angiography before inclusion into the study. None of the patients had exercise-induced ischemia. Electrophysiologic studies and exercise testing were performed on two consecutive days with medications unchanged. All patients had given informed consent to the study.
Digital ECGs including the standard 12 leads of the surface ECG and the orthogonal XYZ Frank leads were recorded for the assessment of T-wave alternans at a sampling rate of 1,000 Hz using a CHI-2000 recording system (Cambridge Heart, Bedford, Massachusetts). The study protocol for comparison of invasive and noninvasive T-wave alternans recordings (19) and the study presented here were both projected prospectively. The recording device has been described in detail previously (19). In brief, high-resolution electrodes were positioned in standard positions after abrasion of the skin surface and testing for low impedance of the electrode location in order to improve ECG quality. After immediate digitization, all data were stored on optical disk for subsequent offline analysis. Selected portions of the continuous recording were exported from the proprietary storage format, downloaded to a Macintosh computer and converted to Labview for Macintosh data format using custom-written software programs (20).
The ECG analysis
The ECG data were then submitted to a previously validated interactive analysis program for ECG repolarization and dispersion variables (10,20). Reproducibility of results with this software program has been shown to be excellent (10,20). The RR interval preceding the single beat chosen for analysis was taken to calculate the exact HR. U waves appeared rarely and were disregarded, using the rules set forth by Lepeschkin (21). If the P wave of the following beat interfered with the T end at faster HRs, the tangent of the T-wave descent was drawn to the baseline. If a biphasic T wave was encountered, the first T peak was measured. Flat T waves with an amplitude of <0.1 mV were considered unmeasurable. Recordings with fewer than 8/12 analyzable leads were excluded from the analysis. If a respective ECG lead was not analyzable at a given HR, it was also excluded at the other HRs in order to ensure comparability of results. Finally, at least five ECGs (at different HRs) had to be analyzable in a single patient. The following ECG variables were prespecified for rate-dependent analysis: QT interval and Q-to-T-peak interval averaged among the analyzable leads, QTD (QTmax − QTmin), precordial QTD (QTmax − QTmin from leads V1 through V6 only), the T-peak-to-T-end (TPE) interval and the area under the T wave (TA)—the latter two also averaged among the analyzable leads.
At the time of electrophysiologic study, antiarrhythmic drugs and beta-blockers had been stopped for at least five half-lives. After placement of right atrial, His bundle and right ventricular catheters, routine electrophysiologic testing was performed, including programmed electrical stimulation from two right ventricular sites with up to three extrastimuli. After induction of sustained ventricular arrhythmias or completion of the protocol, 5 to 10 min were allowed before right atrial pacing for the purpose of this study was begun. First, a recording at the spontaneous sinus rate undisturbed by premature ventricular contractions was completed. Then, right atrial pacing was performed for at least 3 min at a given rate and increasing in steps of 10/min until the AV Wenckebach point or a maximum rate of 130/min was reached.
First, an ECG at the steady-state sinus HR was recorded. Next, submaximal exercise testing was performed, aiming for 70% of the age-specific target HR or a maximum HR of 120 beats/min and with gradually increasing workload. From the continuous digital recording, ECGs at actual HRs of 80, 90, 100, 110 and 120 beats/min were exported to the interactive ECG analysis as described above.
With data divided into invasive and noninvasive subsets, linear regression analysis was done calculating Pearson’s correlation coefficients, intercepts of the regression line, and the slope of the regression line, correcting for repeated measurements in single patients, as described by Glantz and Slinker (22). In addition, a repeated measures analysis of variance (ANOVA) with HR and mode of HR increase as factors was calculated. A p value <0.05 was considered statistically significant.
The 22 men and 2 women patients had an average age of 59 ± 12 years. The underlying cardiac disease was coronary artery disease in 21/24 (87%) patients, dilated cardiomyopathy in 2/24 (8%) patients, and no documented cardiac disease in one patient. Left ventricular function was depressed with an average ejection fraction of 42 ± 15%. Two patients exhibited left bundle branch block. Twelve patients had presented with documented sustained ventricular tachycardia, eight patients with ventricular fibrillation, and another four with both ventricular tachycardia and ventricular fibrillation. Three patients presented with suspected arrhythmogenic syncope. Clinical characteristics of the overall 35 patients were not different from the 24 selected.
Eleven patients were excluded from the analysis because in eight patients fewer than eight ECG leads could be analyzed at all HRs and in three patients the HR increase under exercise was insufficient. Overall, 193 ECG recordings were available from 24 patients. The range of HRs included was 78 ± 14 (minimum) to 116 ± 14 (maximum) during atrial pacing, and 76 ± 17 to 110 ± 17 upon exercise testing. From the invasively recorded ECGs, 10.4 ± 1.2 leads were analyzable; from the exercise ECG recordings, the number of ECG leads was 10.1 ± 1.1.
ECG variables of repolarization
Figure 1A shows all 193 data points for the relationship between Q-to-T-peak interval duration and HR. The r-value and slopes for the linear regression line were calculated separately for exercise testing and atrial pacing data, respectively. A clear linear relationship to HR was found with significant r-values of 0.93, both during exercise testing and atrial pacing (both p < 0.0001). Similar results were found for the rate dependence of the QT interval (Fig. 1B): the R values were 0.93 and 0.93 for exercise testing and atrial pacing, respectively. The corresponding p values were <0.0001 for both groups. By means of repeated measures of ANOVA, HR was found to significantly influence QT and Q-to-T peak (both p < 0.0001), whereas the mode of HR increase was not a significant factor (p = ns).
ECG variables of DVR
All data points for the variables measuring DVR (QTD, precordial QTD, TPE interval, TA) are plotted against HR in Figure 2. By contrast to the variables of repolarization described above, no relationship to HR was found for the variables of DVR: None of the r-values for the linear correlations of QTD reached statistical significance (Fig. 2). Accordingly, the slopes of the calculated regression line ranged close to zero, with values between 0.13 and −0.14, while repeated measures ANOVA found neither HR nor mode of HR increase to be modifiers of any of the four studied variables of DVR.
Rate-dependence of DVR in the human heart
Results of the present study in the human heart demonstrate that DVR is independent of HR over a wide range of physiological HRs. The absence of a rate-dependence of DVR was found for two modes of acceleration of HR, namely an HR increase by exercise as opposed to atrial pacing. The rate-dependent behavior of the three variables of DVR chosen for the study was remarkably similar, although averaging the TPE interval and TA among the 12 leads constitutes a different approach of measuring DVR from the surface ECG than the standard QTD, which is determined by the two extreme QT intervals. The TA and the TPE interval had been validated by us experimentally (20), while the latter variable was recently also introduced as a measure of the transmural DVR by Antzelevitch and co-workers (23).
Potential modifiers of rate-dependence of DVR
The present study confirms our initial hypothesis that DVR is independent of HR, which we had derived from the results of our experimental study (18). Those results could not be directly extrapolated to the human heart, because of the lack of autonomic innervation and the absence of circulating catecholamines in the isolated heart model. Indeed, the two modes of HR modification in the present study were chosen to rule out a confounding effect of sympathetic activation, which can be expected to increase during exercise but to remain unchanged during atrial pacing. In the present study, no additional QT shortening with fast HRs was observed with exercise compared with atrial pacing (i.e., a higher slope of the QT-RR regression line was not detected). Although an effect of sympathetic stimulation on DVR has been demonstrated in a dog model (24), the amount of sympathetic tone was much higher in the experimental situation compared with the difference between pacing and exercise in our study. Clearly, an influence of sympathetic tone or circulating catecholamines on the relationship of HR and DVR was not observed in this study.
The main finding of the present study contrasts with the well-known rate adaptation of the QT interval (13–17) measuring the duration of repolarization, which was confirmed in this study. The rate adaptation of the QT interval has its physiologic basis in the rate adaptation characteristics of action-potential duration, which have been described by several investigators (25–27). It cannot be ruled out that more extreme changes of HR such as bradycardia <50 beats/min or tachycardia >150 beats/min may lead to a change in DVR; however, the study design did not permit us to test such conditions. Similarly, the results may not apply to ventricular premature beats with various coupling intervals, pauses, or short-long-short sequences. Moreover, it should be mentioned that QTD and related ECG variables—as in this study—are accepted as surrogates of DVR, but the correlation between the ECG and DVR may not be perfect (20,28).
Rate-dependence of QTD—previous studies
For comparison in the human heart, the study by Demolis et al. (29) investigated the rate-dependent effects of the class-III agent dofetilide in healthy subjects. They measured QTD defined as the range of measurable QT intervals and did not find any change of QTD during an HR increase under baseline exercise testing. Rate-dependent changes in DVR or QTD may well be present under the influence of action-potential prolonging agents. This has been demonstrated in the human study by Demolis et al. (29) but also in an experimental study from our laboratory (30) studying the rate-dependent effects of various class-III antiarrhythmic drugs on DVR. The present study is the first to investigate rate dependence of QTD in patients with cardiac disease referred for arrhythmia evaluation.
The findings in healthy subjects and patients with cardiac disease in this study are remarkably consistent. Although the results cannot automatically be generalized to all patient populations, such as patients with the congenital long QT syndrome or subjects with the acquired long QT syndrome (29), the body of evidence showing that QTD and similar variables are independent of rate suggests that the common practice of reporting QTc dispersion values (2–7) should not be continued. Without a proper physiological basis the Bazett formula (13) had been used to correct QTD for HR. A recent editorial by Malik et al. (12) had unusually demonstrated that rate-correction of QTD has the potential of producing completely misleading conclusions. From a physiological point of view it does in fact make sense that DVR remains the same value when HR is changed. The rate-adaptation of action-potential duration would be expected to be similar at several locations of the heart. Thus, when the absolute change in duration is not different between the respective sites, the value of the difference—which is DVR—will be expected to remain constant. Fortunately, clinical studies reporting rate-corrected QTD values have usually found similar results for uncorrected QTD and rate-corrected QTD (2–7); therefore, this practice has not influenced the conclusions drawn from these studies.
As measured by QTD or related ECG variables, DVR is not rate-dependent over a wide range of HRs during atrial pacing and exercise testing. The results suggest that the current practice of reporting rate-corrected dispersion variables is unnecessary.
The authors thank Paul Albrecht, PhD, for providing his expert help in the data acquisition and export process, and Birgit Auth for her help during data analysis.
☆ Supported by a grant from the Deutsche Forschungsgemeinschaft, Bonn, Germany (Za 210/1-1).
- dispersion of ventricular repolarization
- heart rate
- QT dispersion
- area under the T wave
- Received January 12, 1999.
- Revision received April 21, 2000.
- Accepted June 26, 2000.
- American College of Cardiology
- Day C.P.,
- McComb J.M.,
- Campbell R.W.F.
- Priori S.G.,
- Napolitano C.,
- Diehl L.,
- Schwartz P.J.
- Hii J.T.Y.,
- Wyse G.D.,
- Gillis A.M.,
- Duff H.J.,
- Solylo M.A.,
- Mitchell L.B.
- Perkiömäki J.S.,
- Koistinen M.J.,
- Yli-Mayry S.,
- Huikuri H.V.
- Zabel M.,
- Klingenheben T.,
- Franz M.R.,
- Hohnloser S.H.
- Coumel P.,
- Maison-Blanche P.,
- Badilini F.
- Bazett H.C.
- Karjalainen J.,
- Viitasalo M.,
- Mäntäri M.,
- Manninen M.
- Merri M.,
- Moss A.J.,
- Benhorin J.,
- Locati E.H.,
- Alberti M.,
- Badilini F.
- Zabel M.,
- Portnoy S.,
- Franz M.R.
- Lepeschkin E.,
- Surawicz B.
- Glantz S.A.,
- Slinker B.K.
- Antzelevitch C.,
- Nesterenko V.V.,
- Shimizu W.,
- Di Diego J.M.
- Opthof T.,
- Misier A.R.,
- Coronel R.,
- et al.
- Arnold L.,
- Page J.,
- Attwell D.,
- Cannell M.,
- Eisner D.A.
- Demolis J.L.,
- Funck-Brentano C.,
- Ropers J.,
- Ghadanfar M.,
- Nichols D.J.,
- Jaillon P.