M. Sosnowski, Z. Czyż and M. Tendera. 3rd Division of Cardiology, Silesian School of Medicine, Silesian Medical Centre, Katowice, Poland

Transcription

1 Europace (21) 3, 39 4 doi:1.13/eupc.2.144, available online at on HEART RATE VARIABILITY Scatterplots of RR and RT interval variability bring evidence for diverse non-linear dynamics of heart rate and ventricular repolarization duration in coronary heart disease M. Sosnowski, Z. Czyż and M. Tendera 3rd Division of Cardiology, Silesian School of Medicine, Silesian Medical Centre, Katowice, Poland Objective QT interval prolongation and increased spatial QT dispersion are important factors increasing the risk in coronary heart disease. The authors studied the spontaneous beat-to-beat variability of ventricular repolarization (RT intervals) in normal subjects and in patients after myocardial infarction (MI) in order to define the determinants of abnormal temporal dispersion. Methods Seventy-six patients with a history of MI (17 female, 9 male, aged 2 1 years) comprised the study group. Forty-seven patients had preserved left ventricular ejection fraction (EF 4%, MI-A) and 29 patients had left ventricular dysfunction (EF<4%, MI-B). Twenty healthy volunteers (6 female, 14 male, aged 2 years) were included as the control group. An ECG signal of 12 heartbeats was recorded in the supine position. After analogue-to-digital conversion (16 bit, 2 khz), the fiducial points of the R wave and T wave were determined. The RR and RT variability (V) assessed in the time domain as the standard deviations of RR and RT (ms), as well as the coefficients of scatterplots of RR and RT intervals. Results As expected, the standard deviation of RR was significantly reduced in MI patients. The magnitude of RTV in the time domain was similar in the controls and in both subgroups of MI patients. The complexity of heart rate variability (HRV) was slightly, but significantly, reduced in the MI-B group, but not significantly in the MI-A heart group. The complexity of RTV behaved in the opposite manner, being increased in both MI subgroups with the lower mean in the MI-B patients. The different behaviour of HRV and RTV was indicated by the increased ratio of RR/RT coefficients, which reached a significantly greater value in the MI-B group. Conclusion The authors have described different patterns of scatterplot of short-term HRV and RTV in normal subjects, which confirmed that RTV is a less complex phenomenon than HRV. In patients after MI, the complexity of HRV diminishes, while the complexity of RTV increases. These opposing changes are more pronounced in patients with left ventricular dysfunction. A possible prognostic value of this feature is unknown and remains to be elucidated in future prospective studies. (Europace 21; 3: 39 4) 21 The European Society of Cardiology Key Words: Complexity, temporal dispersion, computerised electrocardiograms. Manuscript submitted 14 January 2, revised 29 August 2, and accepted November 2. Correspondence: M. Sosnowski, 3rd Division of Cardiology, Silesian School of Medicine, Silesian Medical Centre, Ziolowa St. 47, 4-63 Katowice, Poland. Introduction Abnormal ventricular repolarization is one of the most important factors contributing to an increased risk of adverse outcome in patients with coronary heart disease, after myocardial infarction (MI) and with congestive /1/139+7 $3./ 21 The European Society of Cardiology

2 4 M. Sosnowski et al. heart failure [1]. The common feature of abnormal electrical recovery is the prolongation of the QT interval duration on the routine ECG. An increased spatial dispersion was identified as another risk factor responsible for the occurrence of malignant ventricular arrhythmia and cardiac death [2] ; however, this link was not confirmed by recent studies [3]. These commonly used methods can help to detect the presence of an arrhythmia substrate, but bring no characteristics of momentto-moment changes in ventricular repolarization. Assessment of dynamic properties of beat-to-beat QT intervals may involve time domain and spectral analytic techniques, which are valuable for assessing averaged variation over specific periods or for demonstrating trends over time [4 8]. Methods that come from the chaos theory and non-linear dynamics have already been successfully used to determine the complexity of heart rate variability (HRV) [9 11]. However, their application for evaluating the QT interval dynamics is sparse [12]. Scatterplot (or Poincare plot) of successive intervals is another simple and commercially available nonlinear method used to characterize the heart rate dynamics [13 1]. There are also few reports on the application of scatterplots for evaluation of QT interval dynamics [16 18]. The objective of this study was to determine the dynamics of spontaneous beat-to-beat variability of ventricular repolarization, quantified as the RT interval (RTV) in normal subjects and in patients after MI, and to evaluate the differences in HRV and RTV dynamics resulting from a myocardial injury. about the purpose and methods of the study, and gave their oral consent. Signal recording and processing The study was performed in the supine position during spontaneous breathing, at 1 12 am, after at least 1 min rest. At the time of the study, patients received their drugs normally, according to clinical indications. The list of medications is shown in Table 1. The method of ECG recording and processing has been described in detail elsewhere [12]. Briefly, the ECG signal was recorded using a computer-assisted amplifier from the X, Y, Z bipolar leads. Silver/silver chloride electrodes (ARBO, Mannheim, Germany) were placed according to the recommendations established for signal averaging [19]. Analogue-to-digital conversion was performed with 16-bit resolution and 2 khz of sampling rate per channel (Kardioassist v.3., Medea Corp., Gliwice, Poland). Five hundred and twelve consecutive sinus beats with an automatically evaluated noise level < 2 mv were chosen for further processing. Fiducial points of the R wave and of the offset of the T wave were automatically determined using the Fourier shift method in the lead in which the amplitude of the T wave was maximal (at least 2 mv) and, simultaneously, in which the noise level was minimal. These criteria allowed the optimal signal-to-noise ratio for the T wave to be obtained. Materials and Methods Study subjects Seventy-six patients with a history of MI comprised the study group. There were 17 women and 9 men, aged 2 1 years (range years). Mean time after acute MI was months and ranged from 3 to 6 months. All patients were in sinus rhythm, without premature ventricular or atrial beats throughout the study. Based on the echocardiographic examination (Sonos 2, Hewlett-Packard, U.S.A.) patients were divided into two subgroups: 47 patients with preserved left ventricular ejection fraction (LVEF 4%, MI-A) and 29 patients with depressed left ventricular function (LVEF<4%, MI-B). Clinical characteristics of both subgroups are presented in Table 1. Twenty healthy volunteers, recruited from medical students, hospital staff or their relatives, were included in this study as the control group. There were six women and 14 men. Their ages ranged from 18 to 49 years (mean 2 years). There were no smokers, excessive coffee drinkers, or drug abusers. The study was performed according to the Declaration of Helsinki. All subjects were informed Data analysis The time series of RR intervals was constructed as absolute beat-to-beat interval duration, while the time series of RT intervals was calculated as consecutive differences from the reference value, which was drawn automatically from the template. The absolute values of the RT interval duration were ignored (Fig. 1). The magnitudes of RR and RT variability were assessed in the time domain as the values of standard deviation of mean RR (SD-RR, ms) or of the reference value attributed RT interval (SD-RT, ms). The scatterplots of RR and RT intervals were constructed, which represent the relationship between a given interval plotted on the x-axis against the next (n+1) interval on the y-axis. In such a way, the complexity of RR and RT intervals through a graphic plane presentation was displayed (Fig. 1). For each scatterplot, the correlation coefficient was calculated to evaluate the density of the plot (RRret and RTret), as well as the ratio RR/RT, which describes the similarity of RR and RT scatterplots. A value of RR/RT that equals 1 indicates a similar complexity of HRV and RTV, RR/ RT<1 shows the HRV to be more complex, while RR/RT>1 marks the RTV to be more complex. Statistical analysis was performed using the commercial statistical package (Statistica for Windows PL, Europace, Vol. 3, January 21

3 Dynamics of ventricular repolarization in CHD 41 Table 1 Clinical characteristics of myocardial infarction subgroups Parameter MI-A (n=47) MI-B (n=29) P-value* Age (years) n.s. Gender (female/male) 1/37 7/22 n.s. Time from the MI n.s. CHD risk factors Cigarette smoking (%) 6 66 n.s. Arterial hypertension (%) 78 n.s. Intolerance of glucose or diabetes (%) n.s. Hypercholesterolaemia (%) 68 n.s. Clinical status NYHA I II < 1 III IV 8 MI site (anterior/inferior) 28/19 2/9 n.s. Q/non-Q 33/14 2/9 n.s. More than one MI (%) n.s. Echocardiographic data LVEF (%) < 1 LVEDD (cm) < 1 LV fractional shortening (%) < 1 Angiographic data Number of stenosed vessels n.s. % no significant stenosis 4 n.s. One-vessel disease 14 Two-vessel disease 4 21 Three or more vessel disease Treatment (%) Beta-blockers n.s. Calcium antagonists 1 42 n.s. ACE inhibitors 44 4 n.s. Diuretics 21 2 < Digoxin 9 33 < 1 Antiarrhythmics n.s. Antiplatelet drugs n.s. Anticoagulants n.s. *U-test, Fisher s exact test or median test was used as appropriate. MI-A, MI patients with preserved LVEF; MI-B, MI patients with LV dysfunction; MI, myocardial infarction; LVEF, left ventricular ejection fraction; LVEDD, left ventricular end diastolic dimension; ACE, angiotensin converting enzyme; CHD, coronary heart disease. StatSoft Poland). Mann-Whitney U-test, Kruskal- Wallis ANOVA for ranks and Fisher s exact test were used for comparisons. The differences were considered to be significant with a P value <. Results Means 1 SD values of HRV and RTV in the control group and in the MI subgroups are presented in Table 2. Both normal subjects and MI patients showed similar mean RR intervals, while total HRV, expressed as SD-RR, was significantly reduced in MI patients. The magnitude of RTV in the time domain was similar in the controls and in both subgroups of MI patients, despite highly significant differences in age and SD-RR. The complexity of HRV was slightly, but significantly, reduced in the MI-B group, but not significantly in the MI-A patients. The complexity of RTV behaved in the opposite manner, being increased in both MI subgroups with the lower mean in MI-B patients (Fig. 2). This different behaviour of HRV and RTV was indicated by the increase in RR/RT ratio, which reached a significantly greater value in the MI-B group (Fig. 3). Discussion This study evaluated the variability and complexity of RT intervals and compared them with HRV in normal subjects and in patients after MI. The evaluation of the RTV scatterplot, which described the relation between any given RT interval against the successive interval, provides information about the complexity of the beatto-beat RT interval variation, while a time order is ignored. If the consecutive RT intervals had a similar duration, the scatterplot showed a torpedo-like pattern and the complexity was considered to be low. In contrast, if the differences between consecutive RT intervals were large, then there was a widely dispersed pattern and high complexity was assumed. Europace, Vol. 3, January 21

4 42 M. Sosnowski et al. (a) (b) (c) 11 R = (d) 8 R = Figure 1 An example of heart rate variability and RT interval variability in patients with preserved left ventricular ejection fraction. Upper panel shows the tachograms of (a) RR and (b) RT intervals. Lower panel presents the scatterplot, i.e. the relationship between a given (c) RR or (d) RT. The cloudy pattern of heart rate variability indicates complex dynamics, while the less complex dynamics of RT variability are quantified by a higher value of the correlation coefficient (R). 4 Table 2 Results of heart rate variability and RT interval variability analysis in the study groups Parameter Controls (C) MI-A MI-B P-value* RR interval (ms) n.s. SD-RR (ms) < 1bothvsC SD-RT (ms) n.s. RRret < MI-A vs C < 1 MI-B vs C RTret < MI-A vs C < 1 MI-B vs C < MI-A vs MI-B RR/RT < 1 MI-A vs C < 1 MI-B vs C < MI-A vs MI-B *U-test was used for comparisons. MI-A, myocardial infarction patients with preserved left ventricular ejection fration; MI-B, myocardial infarction patients with left ventricular dysfunction; SD, standard deviation; ret, density of plot (see text). Europace, Vol. 3, January 21

5 Dynamics of ventricular repolarization in CHD 43 (a) (b) (c) R = 776 (d) 1 R = Figure 2 An example of heart rate variability and RT interval variability in patients with depressed left ventricular ejection fraction. Upper panel shows the tachograms of (a) RR and (b) RT intervals. Lower panel presents the scatterplot, i.e. the relationship between a given (c) RR or (d) RT. The pattern of heart rate variability is more dense and indicates less complex dynamics than in patients with preserved systolic function, while the more complex dynamics of the RT interval variability, as quantified by a lower value of the correlation coefficient (R), indicate the increased temporal dispersion of RT variability. 1 The complexity of beat-to-beat ventricular repolarization using the scatterplot has already been studied from direct recordings of the monophasic action potential, where the determination of and 9% repolarization time is relatively easy [17]. However, it seems impractical to use this invasive procedure in a larger population of patients. The present authors evaluated RTV from the surface ECG signal. This was possible by the use of high-resolution recording, which allows detection of the changes in signal amplitude of 13 µv every ms. The authors have found previously that the requirements for the T-wave amplitude and the noise level used in this study allow for reliable determination of the T-wave offset [2]. The patterns of RTV seen in normal subjects form a torpedo-like pattern, in contrast to the widely dispersed points of the HRV scatterplot. This pattern, along with the slightly, but significantly, greater values of the RTV scatterplot s correlation coefficients indicate that the complexity of the RTV is lower than that of HRV. This could be expected, as the phenomenon of the adaptation Controls MI-A MI-B Group Figure 3 Increased RR/RT ratio, indicating greater complexity of short-term RT interval variability as compared with the complexity of heart rate variability, is seen in myocardial infarction patients with depressed left ventricular funtion (MI-B). MI-A, myocardial infarction patients with preserved left ventricular ejection fraction. Points, bars and brackets represent mean SEM ( 1 96 SEM), respectively. RR/RT Europace, Vol. 3, January 21

6 44 M. Sosnowski et al. of ventricular repolarization excludes any significant abrupt changes of QT (here RT) interval duration in a beat-to-beat manner [21]. Therefore, the torpedo-like pattern of the RTV scatterplot is a graphic representation of long-wave control of ventricular repolarization. Several studies using the same method of analysis for evaluation of HRV showed that predominance of long-wave oscillation resulted in the torpedo-like pattern [13 1], similar to that observed in the investigation of Huikuri et al. [17] and the present study of RTV in healthy subjects. As the complexity of HRV is reduced in patients after MI, one could expect a similar behaviour of RTV. Interestingly, scatterplots of RTV in these patients appeared rather more widely dispersed rather than less complex. It may be speculated that such a behaviour of RTV represents a temporal dispersion of ventricular repolarization, as increased spatial dispersion of QT duration has been shown to be a marker of arrhythmia propensity in post-mi patients, in heart failure and in long QT syndromes. Abnormal temporal dynamics of the ST-T segment has already been described as the T-wave alternans. Experimental and clinical data have shown the links between T-wave alternans and vulnerability with lethal arrhythmias [22,23]. As alternans could be considered to be one of several possible patterns of beat-to-beat changes in voltage or duration of the T wave, the RTV could detect both alternans and nonspecifically coupled changes. Thus, the scatterplot of RTV, which allows one, in a simple way, to distinguish visually the RT-interval alternans, also enables one to describe the continuum of different patterns that characterize impaired control of repolarization. A similar conclusion regarding the use of pattern analysis of HRV scatterplots has been previously proposed by Huikuri et al. [14]. The ratio of HRV and RTV scatterplot coefficients may be interpreted as the balance between the control of heart rate and ventricular repolarization. In healthy subjects in the supine postion, the beat-to-beat control of heart rate is dominated by the short-wave oscillations of the vagus, while RT-interval duration is under the control of the long-wave adaptation process. Such a balance is indicated by a ratio of less than 1. In MI patients this ratio is clearly greater than 1, indicating either reduced vagal control, disturbed adaptation process, or both. Interestingly, in MI patients with impaired left ventricular function, this ratio reached the mean value of greater than 3, the main reason being a significant reduction of the RTV scatterplot coefficient. As increased beat-to-beat RTV cannot be explained by the reduction in vagal activity in these patients, the possible explanations are that abnormal haemodynamics of the impaired left ventricle or abnormal respiratory mechanics influence the ventricular repolarization through a mechanical electrical coupling, abnormal filling or altered reflexes [6,24,2]. A mechanical influence of respiration was considered as the main factor for increase in the high-frequency component of RTV by Muller et al., who studied patients with angina pectoris and normal LVEF [6]. This influence might even be more accentuated in those with impaired left ventricles, as shown by Berger et al. [7] ; furthermore, a mechanical electrical transduction was suggested by the experimental study of Stilli et al. [2]. Whether the more complex RTV means it is more chaotic is unanswered since the scatterplot analysis is not the ideal way to detect chaotic phenomena [26]. The present authors have previously shown that fractal dimension of RTV is increased in post-mi patients, especially in those with failing left ventricles [12]. Accordingly, a widely dispersed cloud of points of the RTV scatterplot in post-mi patients with depressed left ventricular function may represent abnormal dynamics of ventricular repolarization. The authors are aware that the coefficient of the scatterplot used is not the optimal descriptor of complex phenomena seen in the graphic display, but the proposed coefficient is easy to compute and understand. Since there is no agreement about how to quantify a scatterplot, and several different ways have been proposed previously [13,14,16,17,27], the authors simply used the coefficient that in non-artifactual recordings is a reliable descriptor of interval distribution. The second limitation of this study may come from the relatively short duration of ECG recordings. It is believed that non-linear processes should be studied from long-term time series, which include thousands of heartbeats [26,27], but the lowest reliable number of beats has not been determined. Huikuri et al. [14] applied quantitative analysis of scattergrams on 1-h ECG recordings to detect the propensity for spontaneous onset of ventricular tachyarrhythmias. The results of this study clearly indicate that the recording of 12 intervals may be enough to distinguish between different patterns found in healthy subjects and in post-mi patients. In conclusion, the authors have described different patterns of scatterplot of short-term HRV and RTV in normal subjects, which confirmed that RTV is a less complex phenomenon than HRV. In patients after MI, the complexity of HRV diminishes, while the complexity of RTV increases. These opposing changes are more obviously expressed in patients with left ventricular dysfunction. The possible prognostic value of this feature is unknown and remains to be elucidated in future prospective studies. References [1] Moss AJ. Measurement of the QT interval and the risk associated with QTc prolongation. A review. Am J Cardiol 1993; 72: 23B B. [2] Day ChP, McComb JM, Campbell RWF. QT dispersion: An indicator of arrhythmia risk in patients with long QT syndrome. Br Heart J 199; 63: [3] Zabel M, Klingenheben T, Franz MR, Hohnloser SH. Assessment of QT dispersion for prediction of mortality or arrhythmic events after myocardial infarction: results of a prospective, long-term follow-up study. Circulation 1998; 97: 243. Europace, Vol. 3, January 21

7 Dynamics of ventricular repolarization in CHD 4 [4] Merri M, Alberti M, Hall WJ, et al. Repolarization duration variability: A tool to quantify the autonomic balance at the ventricle. Comput Cardiol 199; [] Nollo G, Speranza G, Grasso R, et al. Spontaneous beat-tobeat variability of the ventricular repolarization duration. J Electrocardiol 1992; 2: [6] Muller Ch, Dambacher M, Spadacini G, et al. Power spectral changes of spontaneous beat-to-beat variability of RT interval in patients with coronary artery disease. Ann Noninv Electrocardiol 1997; 2: [7] Berger RD, Kasper EK, Baughman KL, et al. Beat-to-beat QT interval variability: a novel evidence for repolarization lability in ischemic and nonischemic dilated cardiomyopathy. Circulation 1997; 96: [8] Theres H, Romberg D, Leuthold T, et al. Autonomic effects of dipyridamole stress testing on frequency distribution of RR and QT interval variability. Pacing Clin Electrophysiol 1998; 11: [9] Denton TA, Diamond GA, Helfant RH, et al. Fascinating rhythm: A primer on chaos theory and its application to cardiology. Am Heart J 199; 12: [1] Goldberger AL. Nonlinear dynamics for clinicians: chaos theory, fractals and complexity at the bedside. Lancet 1996; 347: [11] Lombardi F. Chaos theory, heart rate variability, and arrhythmic mortality. Circulation 2; 11: [12] Sosnowski M, Czyz Z, Petelenz T, et al. Evaluation of nonlinear dynamics of ventricular repolarization in normal subjects and in patients after myocardial infarction. Ann Noninv Electrocardiol 1997; 2: [13] Woo MA, Stevenson WG, Moser DK, Middlekauff HR. Complex heart rate variability and serum norepinephrine levels in patients with advanced heart failure. J Am Coll Cardiol 1994; 23: 6 9. [14] Huikuri HV, Seppanen T, Koistinen M, et al. Abnormalities in beat-to-beat dynamics of heart rate before the spontaneous onset of life-threatening ventricular tachyarrhythmias in patients with prior myocardial infarction. Circulation 1996; 93: [1] Tulppo MP, Makikallio TH, Seppanen T, et al. Heart rate dynamics during accentuated sympathovagal interaction. Am J Physiol 1998; 274: H81 6. [16] Baranowski R, Zebrowski JJ, Poplawska W, et al. 3-dimensional Poincare plots of the QT intervals an approach to nonlinear QT analysis. Comput Cardiol 199; [17] Huikuri HV, Airaksinen KEH, Koistinen MJ, et al. Two-dimensional vector analysis of beat-to-beat dynamics of ventricular repolarization. Ann Noninv Electrocardiol 1997; 2: 121. [18] Sosnowski M, Czyz Z, Petelenz T, et al. Clinical correlations of nonlinear dynamics of heart period and ventricular repolarization in patients after myocardial infarction. In: Liebman J, ed. Electrocardiology 96: From the Cell to the Body Surface. Singapore: World Scientific Publishing Co. 1997: [19] Breithardt G, Cain Me, El-Sherif N, et al. Standards for analysis of ventricular late potentials using high-resolution or signal-averaged electrocardiography: A statement by a Task Force Committee of the European Society of Cardiology, the Americal Heart Association, and the Americal College of Cardiology. J Am Coll Cardiol 1991; 17: [2] Sosnowski M, Czyz Z, Petelenz T. Could we really determine the T-wave duration. An answer comes from T-wave triggered signal averaged ECG. Comput Cardiol 199; [21] Franz MR. Methods and theory of monophasic action potential recordings. Prog Cardiovasc Dis 1991; 33: [22] Nearing BD, Oesterle SN, Verrier RL. Quantification of ischemia induced vulnerability by precordial T wave alternans analysis in dog and human. Cardiovasc Res 1994; 28: [23] Rosenbaum DS, Jackson LE, Smith JM, et al. Electrical alternans and vulnerability to ventricular arrhythmias. N Engl J Med 1994; 33: [24] Taggart P, Sutton P, Lab M. Interaction between ventricular loading and repolarization. Relevance to arrhythmogenesis. Br Heart J 1992; 67: 213. [2] Stilli D, Aimi B, Sqoifo A, et al. Dependence of temporal variability of ventricular recovery on myocardial fibrosis. Role of mechanoelectric feedback? Cardiovasc Res 1998; 37: 8 6. [26] Elbert T, Ray WJ, Kowalik ZJ. Chaos and physiology: deterministic chaos in excitable cell assemblies. Physiol Rev 1994; 74: [27] Hnatkova K, Cope X, Staunton A, Malik M. Numeric processing of Lorenz plot of R-R intervals from long-term ECGs. Comparison with time-domain measures of heart rate variability for risk stratification after myocardial infarction. J Electrocardiol 199; 26(Suppl): Europace, Vol. 3, January 21