Characterization of the non-linear rate-dependency of QT interval in humans

Transcription

1 Europace (2003) 5, doi: /eupc Characterization of the non-linear rate-dependency of QT interval in humans G. Malfatto 1, M. Facchini 1, A. Zaza 2 1 Divisione di Cardiologia, Ospedale San Luca, Istituto Auxologico Italiano IRCCS; 2 Dipartimento di Biotecnologie e Bioscienze, Università di Milano-Bicocca, Milano, Italy Aims Repolarization has rate-dependent and rateindependent components. A function considering such components separately was validated in canine Purkinje fibres and applied to the QT/RR relation in humans. Methods and Results Action potential duration (APD) was measured in Purkinje fibres during steady-state pacing at different cycle lengths (CL) and after prolonged quiescence (APD 0 ). The APD/CL relationship was expressed by this function: APD=APD * max CL S /(CL S 50 +CL S ), where APD max (APD extrapolated at infinite CL) is a rateindependent measure of repolarization, CL 50 (CL at which 50% of APD max is achieved) and S evaluates the rate dependency of APD. The same function was used to fit the QT/RR relation in 46 normal subjects (20 males, 26 females) and in 7 amiodarone-treated subjects undergoing a bicycle stress test. RR and QT (V 5 ) were measured at the end of each load step; QT c (Bazett s formula) was obtained at rest. The APD/CL and QT/RR relations were equally well expressed by the function with high correlation coefficients (R 0 90). In Purkinje fibres, APD max was ms, CL 50 was ms and S was APD max and APD 0 correlated (R=0 96) and were similar. The corresponding values in humans were: QT max ms, RR ms and S While QT c and QT max were longer in females, RR 50 and S were similar between genders. Amiodarone increased QT c,qt max and RR 50 and decreased S. In QT max and QT c distributions generated by pooling data from treated and untreated subjects, 86% of treated subjects were correctly identified by QT max and 28% by QT c. Conclusions Canine and human repolarization showed a saturating dependency on cycle length, described by the proposed function. Gender and amiodarone independently affected QT max,rr 50 and S: therefore they might reflect specific ionic mechanisms. Finally, QT max identified druginduced repolarization abnormalities in individual subjects better than QT c. (Europace 2003; 5: ) 2003 The European Society of Cardiology. Published by Elsevier Science Ltd. All rights reserved. Key Words: Heart rate, QT interval, ventricular repolarization. Introduction The duration of repolarization, evaluated as action potential duration (APD) at the cellular level and as QT interval on the ECG, is strongly affected by heart rate. This is the consequence of the time-dependency of many among the currents underlying repolarization and of rate-related changes in the intra- and extracellular ionic environments [1]. Thus, to characterize repolarization and its abnormalities, a quantitative description of its Manuscript submitted 5 June 2002, and accepted after revision 23 December Correspondence: Dott.ssa Gabriella Malfatto, Divisione di Cardiologia, Ospedale San Luca, Istituto Auxologico Italiano IRCCS, via Spagnoletto, Milano, Italy. Tel.: ; Fax: ; Malfi@Auxologico.it rate-dependency must be provided. To this end, indices have been developed in the past by empirically fitting the relation between QT and RR intervals in man with a variety of mathematical functions (linear, cube-root, mono- or biexponential, etc.) [2 5]. In addition, linear fits of the QT/RR relation have been used to identify the repolarization abnormality in subjects with the idiopathic long QT syndrome [6]. Rate-corrected indices derived from population studies are often inadequate when used to detect repolarization abnormalities in individual subjects, because QT/RR patterns show not only inter-individual, but often intra-individual variability [7,8]. In addition, QT/RR functions proposed so far were described by parameters devoid of an explicit physical meaning, whose changes could not easily be related to the underlying mechanisms. Particularly, Bazett s formula (QT (n) =QT c * RR (n 1) 0 5 ) widely /03/ $35.00/ The European Society of Cardiology. Published by Elsevier Science Ltd. All rights reserved.

2 164 G. Malfatto et al. used in the clinical setting for its simplicity describes the steepness of the QT/RR relation after correcting for its non-linearity. However, the fact that QT c does not consider the possibility that rate-dependent and rateindependent mechanisms of repolarization might be modulated separately, thus restricting its applicability. In the present study we evaluated the possibility to describe human rate-dependency of repolarization with a simple function, derived from previous experimental work [9 12], which has the following features: (1) it is suitable to treat rate-dependency of repolarization and rate-independent duration of repolarization separately; (2) its parameters have an explicit physical meaning. Function parameters were estimated in individual subjects from analysis of QT values over a wide range of RR intervals, produced by graded physical exercise. The function was initially validated on action potential duration measurements from isolated canine Purkinje fibres and then applied to QT measurements in human subjects. Performance of the function parameters in detecting gender-related and drug-induced differences in repolarization was then tested. Methods In vitro study The following investigation conforms with the Guide for the care and use of laboratory animals (U.S. National Institute of Health, NIH publication No 85-23, revised 1996). Four mongrel dogs were anaesthetized with sodium thiopental (50 mg/kg i.v.). Their hearts were rapidly removed through a left thoracotomy and placed in cold Tyrode s solution. Free running Purkinje fibres were dissected from both ventricles, pinned at the bottom of a Lucite chamber and superfused (temperature= C) with Tyrode s solution of the following composition in mm: NaCl 131, KCl 4 0, CaCl 2 2 0, MgCl 2 0 5, NaH 2 PO 4 1 8, NaHCO 3 18, dextrose 5 5 (ph ). Transmembrane potential was recorded, during electrical stimulation, with standard microelectrode techniques. The study protocol was as follows. A 1 h equilibration period was followed by 2 min stimulation runs at cycle lengths (CL) varying between 0 25 and 5 s in random sequence; each run was preceded by 2 min quiescence. Each fibre received at least eight pacing sequences, and some CL were repeated twice to check for stability of the measurement. Action potential duration at 90% repolarization (APD 90 ) was electronically measured off-line with an accuracy of 125 μs. The first action potential elicited after prolonged quiescence (>1 min, approximating infinite CL) was indicated as APD 0. The APD measured during steady state stimulation (end of each 2 min run, average from five cycles), was plotted against stimulation CL and expressed by the relation APD=APD max * CL S /(CL 50 S +CL S ) (1) where APD max is APD at infinite CL (i.e. a rateindependent measure of APD), CL 50 is the CL at which 50% of APD max was achieved (i.e. the CL value around which APD modulation is centered) and S describes the extent (or steepness) of the CL-dependency of APD. The parameter S was initially set to 1 as in previous studies [11 12] ; however, after finding that QT/RR relations in man could not be adequately expressed where S=1, the procedure was repeated by introducing S as an independently estimated parameter. Thus, when S=1 as in vitro, the relationship equals a hyperbolic, Michaelis Menten type function such as the one previously published (where APD max and CL 50 are analogous to V max and K m respectively) [11,12]. The larger S, found in humans, implies a steeper dependency of repolarization on CL, i.e. one occurring over a narrower range of CL values. Clinical study We analysed the ECG of 53 subjects (26 males and 27 females) undergoing a symptom-limited bicycle stress test as a routine evaluation for sports practice. The investigation was approved by our Ethical Committee, and conformed with the principles outlined in the declaration of Helsinki. Of these patients, 7 (6 males, 1 female) were taking amiodarone as antiarrhythmic prophylaxis of lone paroxysmal atrial fibrillation. All subjects had a normal physical examination, normal serum electrolytes, normal ECG, normal M-mode and 2D echocardiogram. The test used a step protocol increasing the load by 25W every 2 min (Marquette CASE 16, Marquette Inc, Milwakee, WI, U.S.A.). The duration of each step was sufficient for heart rate to approximate a steady state; at least seven, but most of the time 9 to 10 protocol steps (including baseline and recovery) were performed in each patient. The tests were stored on a diskette and analysed manually off-line. For any level of exercise, the RR and QT intervals of the last 15 s in lead V 5 were measured with an electronic pen (tolerance of 1 ms) and averaged. If a U wave was present, the end of the T wave was taken at the deepest point of the QTU deflection. The same QT intervals were blindly measured by a second investigator and, whenever the difference between the two measurements exceeded 10 ms, a third blind reading was obtained. The various readings were then averaged to obtain a final QT value. Values of QT were then plotted against the corresponding RR interval and expressed by the following relation: QT=QT max * RR S /(RR 50 S +RR S ) (2) where QT max,rr 50 and S have the same meaning as the corresponding parameters of equation 1. A ratecorrected QT value (QT c ) was also computed on the baseline ECG, as a part of the routine examination, according to the Bazett s formula: QTc=QT (n) /RR 0 5 (n-1). We choose to measure QT from V 5 because in this lead

3 QT/RR relationship in healthy subjects 165 Figure 1 APD/CL relation in canine Purkinje fibres. Panel A: an example of APD/CL relation and its interpolation by equation 1 (solid line). Panel B: plot of APD max vs APD 0 estimated and measured from the same fibres respectively. The solid line represents a linear fit of the data (parameters in the text). the T wave was often sharper and easier to measure; it was beyond the purpose of this study to discriminate whether the results could differ according to lead selected for QT measurement. Curve fitting and statistical analysis The APD/CL and QT/RR relationships were expressed by equations 1 and 2, respectively, by using a non-linear, least square fitting routine (Origin 6.0, Microcal, U.S.A.). The fitting correlation coefficients were 0 9 and highly significant in all cases. Results are expressed as mean value SD. Differences between means were tested by Student s t-test. Correlation between variables was tested by linear regression and expressed as the correlation coefficient (R). Significance of correlation in linear and non-linear fitting was tested by χ 2 analysis. Normality of distribution was tested by Kolmogorov Smirnov statistics (K-S). A P<0 05 was considered significant. Results In vitro study A total of 14 Purkinje fibres was studied. As shown in Fig. 1A, the APD/CL relation was curvilinear, tending to an asymptotic value at longer CL. Data points were accurately fitted by equation 1 (R>0 90) to obtain APD max, CL 50 and S estimates (Fig. 1, panel A). APD max was ms, CL 50 was ms and S was To test the predictive value of equation 1, APD max was compared with APD 0 in each fibre (Fig. 1, panel B). Mean values of APD 0 (469 49) and APD max were similar and measurements of the two parameters were linearly related (R=0 96; P<0 05) with a regression coefficient of (ns vs 1 0). Clinical study A total of 20 males (age years, range 15 60) and 26 females (age years, range 14 71; ns vs males) not taking drugs completed the exercise stress test. The maximum workload was W for males and W for females (P<0 05). The heart rate at peak exercise was bpm for males and bpm for females (ns), exceeding 85% of the maximal predicted value in both groups. Working capacity and coronary reserve were normal in all subjects. A further group of patients (6 males and 1 female, mean age 53 5 years) was under treatment with amiodarone at a dose of 200 mg/day to prevent recurrence of paroxysmal atrial fibrillation, and was separately analysed. Although the range of RR intervals that could be evaluated was considerably narrower than in the in vitro experiments, a curvilinear QT/RR relation with tendency to saturate at long RR intervals could be clearly observed in most cases. Examples of individual QT/RR relations are shown in Fig. 2. In the 46 untreated subjects, the steady-state QT/RR relation was accurately fitted by equation 2, with an average correlation coefficient of (P<0 05). The mean values of the estimated parameters were QTc , QT max ms, RR ms and S Figure 3 shows individual QT values, pooled from all the 46 subjects, plotted against the corresponding RR interval. Fitting of pooled values with equation 2 led to QT max, RR 50 and S estimates very close to the mean of values

4 166 G. Malfatto et al. Figure 2 Examples of QT/RR relationships in a healthy male (panel A) and female (Panel B) and in an amiodarone-treated male (panel C). The solid lines represent fitting of the data points with equation 2; values of the estimated parameters are reported in the inset. Figure 3 QT/RR relationship in the study population of 46 healthy subjects (pooled data). All the values of QT were plotted against the corresponding RR interval. The solid line represents the fit of pooled data by equation 2 (parameters in the text). obtained from individual curves. The tendency of the curve to plateau at longer RR intervals is evident. Separate analysis of the parameters in males, females and amiodarone-treated subjects is shown in Fig. 4. Compared with males, females had longer QT c ( vs ; P<0 05), longer QT max ( ms vs ms, P<0 05), but similar values of RR 50 ( ms vs ms, ns) and S ( vs ms, ns). QT max was not correlated with basal heart rate (R= 0 17; ns) and, within each gender group, the values of all parameters were independent of age. In the 7 subjects treated with amiodarone (an example is shown in Fig. 2, panel C), QT c was , QT max was ms, RR 50 was ms and S was (all values, P<0 05 vs untreated subjects). Values of QT c and QT max measured from the entire sample (including amiodarone treated subjects) were significantly, but loosely correlated (R=0 48; P<0 05); such correlation was lost when only untreated subjects were considered (R=0 17; ns). It was noteworthy to compare QT max and QT c for their ability to spot individual cases with prolonged repolarization within a sample. To this end, distributions for QT c and QT max values from the entire sample (untreated+treated subjects) were generated (Fig. 5); amiodarone treatment was assumed to prolong repolarization. The QT c distribution slightly deviated from normality (K-S statistics=0 132; P=0 022) and was expressed by a single gaussian function with a high correlation coefficient (R=0 86; P<0 05) (Fig. 5, panel a). The QT max distribution strongly deviated from normality (K-S statistics=0 211; P<0 0009) and could be expressed only by the sum of two gaussian curves (R=0 93; P<0 05) (Fig. 5, panel b). The values corresponding to the 95th percentiles for the QT c (461) and for the lower QT max distribution (485 ms) were calculated from the respective cumulative probability curves (see insets in panels a and b of Fig. 5). The sharper clustering of QT max values can also be appreciated from a plot of QT max vs QT c values (Fig. 6). By adopting the 95th percentile of each parameter as a cutoff value to define normality (solid lines in Fig. 6), it can be appreciated that only 2 of the 7 amiodarone treated patients (28%, open symbols) were identified by QT c, while 6 of 7 (86%) were identified by QT max (P<0 05). it should be noticed that in 3 of 46 untreated subjects (6%, all females) QT max exceeded the 95th percentile. Among subjects with normal QT max (<485 ms), RR50 and S showed a significant, but loose correlation with QT max (RR 50 R=0 58, P<0 05; S: R= 0 49; P<0 05). In identifying amiodarone-treated patients RR 50 was as efficient as QT max, but did not show any clear cut advantage over it.

5 QT/RR relationship in healthy subjects 167 Figure 4 Average values of QT c and of the parameters estimated from the QT/RR relation in males (n=20), females (n=26) and amiodarone-treated subjects (amio, n=7). Symbols represent mean SD. *=P<0.05 (amio is compared to pooled untreated subjects). Discussion In this study we identified a function which interpolated the steady-state APD/CL relation of canine Purkinje fibres, accurately predicting their action potential duration at infinite CL. The same function also fitted the QT/RR relation obtained under quasi steady-state conditions in normal subjects during exercise stress testing. The parameters describing the QT/RR relation in humans were differentially affected by conditions (gender and amiodarone therapy) associated with an otherwise indistinguishable prolongation of QT c. Moreover, within the population studied, QT max appeared to have a larger sensitivity than QT c in identifying amiodaroneinduced repolarization abnormality. APD/CL vs QT/RR relations As proven by its close correlation with APD 0, APD max is a reliable estimate of action potential duration at infinite cycle length. Cycle lengths long enough to be considered as infinite cannot be practically achieved in man; nonetheless, the QT/RR relationship showed a clear cut tendency to plateau in most cases, as shown by Figs 3 to 5. Under such conditions, QT max may provide a reasonable estimate of QT at infinite RR intervals. Compared with the APD/CL relation of Purkinje fibres, the human QT/RR relation had slightly smaller QT max and RR 50 and larger S values. The finding that S values larger than 1 were required to fit accurately the QT/RR relation of patients was actually the motive for the introduction of the parameter S, fixed at 1 in previous experimental characterizations of the APD/CL relation [10 12]. A larger S implies that the relation between repolarization and RR intervals was steeper, and the range of CLs encompassing it was narrower in humans than in canine Purkinje fibres. The conditions under which the relationship between repolarization length and cycle length could be obtained in experimental and clinical studies were clearly different. Most notably, while in patients the increase in sympathetic activity, reflected by RR interval shortening, might contribute to, or modulate the attending changes in QT, in isolated tissues the adaptation of action potential duration could be safely and only attributed to cycle length changes. Second, while action potential duration was measured in vitro from a single type of tissue (i.e. the conduction system), regional differences in the ionic mechanisms underlying repolarization between Purkinje fibres and working myocardium [13], and an unequal contribution of adrenergic modulation [14] contribute to the genesis of the QT interval, and thus may explain the discrepancies observed. Finally, true steady-state conditions can be achieved in vitro, but can only be approximated in vivo. Although the dissimilarities between the conditions of the measurements need to be considered in the

6 168 G. Malfatto et al. Figure 5 Distributions of QT c (a) and QT max (b) values from the entire sample (treated+untreated). The best fitting function (solid line) was a single gaussian distribution in the case of QT c and the sum of 2 gaussian curves for QT max. The insets show the cumulative probability plots computed from the respective functions; the arrows mark the QT values corresponding to the upper 95th percentile of each distribution. implies that, if future experimental studies will succeed in associating each function parameter to a sufficiently specific set of ionic mechanisms, the QT/RR relation may provide a relatively simple way to infer the basic defect underlying repolarization abnormalities in patients. Comparison between the parameters of the QT/RR relation and QT c Figure 6 Relationship between QT max and QT c the entire sample (solid symbols: untreated subjects; open symbols amiodarone-treated subjects). The dotted and solid lines represent the upper 95th percentiles of QT c and QT max distributions respectively. interpretation of results, the parameters estimated by fit of data with the proposed function might retain a similar meaning in the experimental and clinical settings. This As recently suggested [7], rate-corrected indices may be more efficient in detecting repolarization abnormalities if compared within the specific sample tested. Accordingly, the term normal will be used here to define parameter values falling within the 95th percentile of the distributions identified in the present study; i.e. without reference to previously defined normal values. When examined among untreated subjects with normal values, QT max and QT c were not significantly correlated. On the other hand, both QT c and QT max were able to detect gender-related and drug-induced differences in the duration of repolarization. This suggests that both QT c and QT max are representative of repolarization, but they may carry different information. Although mean QT c value was significantly prolonged by amiodarone, there was substantial overlap between the data from treated and

7 QT/RR relationship in healthy subjects 169 untreated patients, thus resulting in an apparently unimodal distribution and a low sensitivity in identifying repolarization abnormality in individual cases. On the other hand, QT max distribution was bimodal, with 6 of 7 amiodarone-treated patients belonging to a distinct subset of values. Thus, compared with QT c,qt max appears to be much more sensitive in detecting amiodaroneinduced repolarization changes in individual subjects. Three of 46 untreated subjects with a normal QTc had an abnormally long QT max, thus potentially questioning the specificity of QT max. In one of these cases, the QT/RR relation had relatively little tendency to saturate, compared with the rest of the population, thus potentially leading to errors in the estimation of QT max. Albeit a previously undetected repolarization abnormality cannot be ruled out, an explanation could not be found for the remaining two cases. The other parameters of the QT/RR relation, RR 50 and S, were loosely but significantly correlated with QT max. This might reflect either a biological relation, or simply a redundancy in the parameters describing the QT/RR function. However, while amiodarone-induced changes in QT max were associated with RR 50 and S changes, gender-related ones were not. Albeit not conclusive, this observation supports the view of a biological cause for the correlation. A similarity of RR 50 and S values between males and females may be interpreted to suggest that the ratedependency of repolarization is similar between genders. This might appear in contrast to previous observations showing that the slope of linear interpolation of the QT/RR relation may be steeper in females [15]. However, it should be considered that linear interpolation is not suitable to discriminate rate-dependent aspects of repolarization from rate-independent ones (e.g. a steeper linear QT/RR relation may result from an increase in QT max alone); thus the conflict is only apparent. Moreover, even if it is known that female gonadal steroids decrease the level of expression of delayed rectifier channels in cardiac cells [16] (a potential mechanism of gender-related differences in repolarization), it is unknown whether such changes may affect the intrinsic duration of repolarization, its rate-dependency or both. Limitations of the study Equation 1 has the form of the Hill equation, commonly used in biochemistry to define the relation between reaction velocity and substrate concentration. Accordingly the parameters QT max and CL 50 and S may be considered equivalent to Hill s V max7 K m and n. While the Hill equation was rigorously derived from first order reaction kinetics, it is unknown whether such a model may also apply to the rate-dependency of repolarization. Thus, the use of a Hill-type function in the present study is purely empirical and has the only purpose of providing a quantitative estimate of the QT/RR relation. Estimation of the QT/RR parameters requires extrapolation. Thus, the estimates might be biased if, over the range of RR intervals that can be examined in an individual patient, QT values do not show a clear tendency to plateau. According to the present results, satisfactory QT/RR relations (see figures) can be obtained in most subjects; nonetheless, if such condition were not verified, the results should be regarded with caution. An alternative to exercise testing in QT/RR sampling, might be provided by 24 h Holter recordings, a method previously used to obtain individual QT/RR patterns [7,8]. Although we did not directly compare the performance of the two methods, several considerations suggest that exercise testing may be preferable. First, the wide range of RR intervals achieved during controlled exercise is seldom available in Holter recordings. Second, beat-to-beat variability of RR intervals is usually decreased during steady exercise, thus allowing a closer approximation to steady-state conditions. On the other hand, depending on the degree of RR variability, the QT/RR relation during Holter recordings would probably be intermediate between the one representative of steady-state and that typical of restitution [9 12], with the additional influence of the lag hysteresis of the QT interval [17]. Conclusions Evaluation of the QT/RR relation by a function allowing independent estimates of the rate-dependency of repolarization (RR 50 and S) and of its rate-independent duration (QT max ) is able correctly to detect repolarization abnormalities within a sample (i.e. in the absence of an internal reference value). The information provided by QT max,rr 50 and S is partly consistent with the one obtained from QT c, but does not overlap with it and might be more specific. In particular, QT max may be more suitable than QT c in detecting drug-induced repolarization abnormalities. A pathophysiological interpretation of RR 50 and S indices requires further information from specifically designed studies. Nonetheless, the introduction of such indices allows QT max to be a truly rate-independent estimate of repolarization and might explain its better performance over QT c. Finally, the non-linearity of the QT/RR relation implies that linear fits of the QT/RR relation should not be used to compare the rate-dependency of repolarization among patients samples with different basal heart rates. This study was partially supported by a grant COFIN-MURST 2000 to A.Z. References [1] Boyett MR, Jewell BR. Analysis of the effects of changes in rate and rhythm upon electrical activity in the heart. Prog Biophys Molec Biol 1980; 36: [2] Kligfield P, Lax KG, Okin PM. QT interval-heart rate relation during exercise in normal men and women: definition by linear regression analysis. J Am Coll Cardiol 1996; 28: [3] Ahnve S. Correction of the QT interval for heart rate: review of different formulas and the use of Bazett s formula in myocardial infarction. Am Heart J 1985; 109:

8 170 G. Malfatto et al. [4] Karjalainen J, Viitasalo M, Manttari M, et al. Relation between QT intervals and heart rates from 40 to 120 beats/min in rest electrocardiograms of men and a simple method to adjust QT interval values. J Am Coll Cardiol 1994; 23: [5] Aytemir K, Maarouf N, Gallagher MM, et al. Comparison of formulae for heart rate correction of QT interval in exercise electrocardiograms. Pacing Clin Electrophysiol 1999; 22: [6] Merri M, Moss AJ, Benhorin J, et al. Relation between ventricular repolarization duration and cardiac cycle length during 24-hour Holter recordings. Findings in normal patients and patients with long QT syndrome. Circulation 1992; 85: [7] Malik M. Problems of heart rate correction in assessment of drug-induced QT interval prolongation. J Cardiovasc Electrophysiol 2001; 12: [8] Batchvarov V, Malik M. Individual pattern of QT/RR relationship. Cardiac Electro-Physiological Review 2002; 6: [9] Elharrar V, Surawicz B. Cycle length effect on restitution of action potential duration in dog cardiac fibers. Am J Physiol 1983; 244: H [10] Zaza A, Malfatto G, Schwartz PJ. Sympathetic modulation of the relation between ventricular repolarization and cycle length. Circ Res 1991; 68: [11] Malfatto G, Zaza A, Schwartz PJ. Parasympathetic control of cycle length dependence of endocardial ventricular repolarisation in the intact feline heart during steady state conditions. Cardiovasc Res 1993; 27: [12] Malfatto G, Zaza A, Vanoli E, Schwartz PJ. Muscarinic effects on action potential duration and its rate dependence in canine Purkinje fibers. Pacing Clin Electrophysiol 1996; 19: [13] Antzelevitch C, Sicouri S, Litovsky SH, Lukas A, Krishnan SC, Di Diego JM, Gintant GA, Liu DW. Heterogeneity within the ventricular wall. Electrophysiology and pharmacology of epicardial, endocardial, and M cells. Circ Res 1991; 69: [14] Burashnikov A, Antzelevitch C. Block of I(Ks) does not induce early after-depolarization activity but promotes beta-adrenergic agonist-induced delayed afterdepolarization activity. J Cardiovasc Electrophysiol 2000; 11: [15] Stramba-Badiale M, Locati EH, Martinelli A, et al. Gender and the relationship between ventricular repolarization and cardiac cycle length during 24-h Holter recordings. Eur Heart J 1997; 18: [16] 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: [17] Lau CP, Freedman AR, Fleming S, Malik M, Camm AJ, Ward DE. Hysteresis of the ventricular paced QT interval in response to abrupt changes in pacing rate. Cardiovasc Res 1988; 22: