Origin on the electrocardiogram of U-waves and abnormal U-wave inversion

Transcription

1 Cardiovascular Research 53 (2002) locate/ cardiores Origin on the electrocardiogram of U-waves and abnormal U-wave inversion Diego di Bernardo, Alan Murray* Regional Medical Physics Department, Freeman Hospital, Newcastle-upon-Tyne NE7 7DN, UK Received 12 June 2001; accepted 2 August 2001 Abstract Aims: Soon after the initial development of electrocardiography, U-waves were discovered in many normal subjects following the T-wave repolarisation waveform on the electrocardiogram. Various explanations have been offered for their origin, but none is universally accepted. We used our model of left ventricular repolarisation to explore the most common hypotheses for the genesis of U-waves. Methods: Recently, we have shown that a computer model of left ventricular repolarisation was able to explain the formation of the characteristic shape of the T-wave, and we have now used this model to explore the most common hypotheses for the genesis of U-waves. The repolarisation phase of the action potentials in the model exhibited an after-potential. We investigated separately the effect on the 12-lead electrocardiogram of three different features of the model: the amplitude of the after-potential; dispersion of repolarisation in the left ventricle ranging from 20 to 100 ms; the timing of the after-potential, relative to the end of the principal action potential component, ranging from 2100 to 100 ms. Results: We show that delaying repolarisation in different regions of the heart cannot explain the U-wave, but show that the presence of after-potentials on the cardiac action potential do explain the U-wave polarity and other characteristic U-wave features. We also show that abnormal after-potential timing corresponds with abnormal U-wave inversion. Conclusion: Our model provides a realistic and simple solution to the problem of U-wave genesis Elsevier Science B.V. All rights reserved. Keywords: Computer modelling; Conduction system; ECG; Membrane potential; Repolarization; Stretch/ m e coupling 1. Introduction genesis were reviewed by Surawicz [2]. There are three main hypotheses: late repolarisation of the Purkinje fibers In the normal electrocardiogram (ECG), an extra deflec- [1], late repolarisation of some other portions of the left tion at the end of the obvious repolarisation sequence is ventricle, and alterations in the normal action potential often seen. This U-wave, as it was named by Einthoven in shape by after-potentials, which are most likely generated 1903 [1], is still the subject of debate almost 100 years by mechano-electric feedback [4]. later, as its origin is still uncertain and the various Involvement of the Purkinje fibres is the oldest proposal hypotheses underlying its origin are controversial [2]. and is thought to be the least likely explanation. Lepesch- The U-wave in normal subjects always has the same kin [4] and Surawicz [2] listed many observations that polarity as the T-wave, and so when the U-wave inverts make it difficult, if not impossible, to reconcile this with respect to the T-wave this is of diagnostic importance proposal with the formation of U-waves. Late repolarisa- [3]. Currently, however, the main reason for studying the tion of some other portions of the left ventricle was first U-wave is to gain an insight into the underlying cardiac proposed by Einthoven [1], and more recently by Anelectrophysiology, as well as simply to resolve the specula- tzelevitch and Sicouri [5] who attributed U-waves to the tion surrounding its origin. The hypotheses for U-wave late repolarisation of M cells, found in the mid-myocardium. Lazzara [6] presented many arguments against the *Corresponding author. Tel.: ; fax: involvement of M cells by demonstrating that these cells address: alan.murray@ncl.ac.uk (A. Murray). Time for primary review: 29 days / 02/ $ see front matter 2002 Elsevier Science B.V. All rights reserved. PII: S (01)

2 prolong the T-wave rather than create distinct U-waves. In addition, recent experiments suggest that M cell action potential duration in vivo is not substantially different from those of normal myocardial cells [7 9]. Currently, the hypothesis involving after-potentials is most favoured [2], as it accords with the observation that mechanical events associated with ventricular wall motion are correlated with both U-waves on the surface ECG and also with after-potentials on the cardiac cellular action potential waveform [2,10]. Franz et al. [11] recorded afterpotentials while mechanically stressing cardiac cells in an in vitro preparation. Cellular electrophysiologists have also discovered ionic transfer associated with these after-po- tentials [12]. However, firmer proof of the link between after-potentials and U-waves appearing on the body-sur- face ECG is still lacking. D. di Bernardo, A. Murray / Cardiovascular Research 53 (2002) hence have been able to examine the after-potential hypothesis for U-wave genesis. The hypothesis associated with late repolarisation in some portions of the left ventricle was also studied, by changing dispersion of repolarisation. In addition, we were able to research the effect of different delays between the normal action potential phases and the timing of the after-potential, which allowed us to investigate whether abnormal delays produced abnormal U-waves Model of action potentials The repolarisation phase of the action potentials used in the model exhibited an after-potential. The repolarisation phase waveform chosen was similar to the one described experimentally by Franz et al. [11] during mechanical stretch of the myocardial cell. This was obtained as the sum of two components: 2. Methods 1. A normal action potential repolarisation phase, shown in Fig. 1 (middle panels), which was described mathe- We have already proposed a computational model of left matically as a product of exponential functions [17], ventricular repolarisation [13], which successfully exwhose parameters determine the shape and duration of plained the formation and shape of the T-wave [14]. The the action potential. model is described in the publications cited [13,14]. It uses 2. An after-potential component (APC), shown in Fig. 1 an experimentally derived repolarisation sequence in a (top panel), described as a product of exponential three-dimensional left ventricle, embedded in a three-difunctions whose parameters determine its shape and mensional torso, to produce 12-lead ECG T-waves. The duration. torso surface was represented by a cylinder 50 cm tall with an elliptic cross-section (major diameter 34 cm, minor By varying the magnitude of the APC, i.e. multiplying it diameter 26 cm). It was considered electrically homoge- by a constant scaling factor, we were able to obtain action neous and isotropic with a conductivity equal to 0.3 S/ m. potentials exhibiting different after-potential characteristics The left ventricle was modelled by a truncated ellipsoid (Fig. 1, bottom panel). To quantify the amount of afterrepresenting the epicardium (height 10.6 cm truncated at potential in an action potential, we used the H-ratio defined 7.0 cm, major diameter 7.0 cm, minor diameter 6.6 cm). by Coraboeuf et al. [12]. The H-ratio measures the An approximately ellipsoidal cavity represented the endo- decrease in the slope at the end of the action potential cardium. The left ventricle (LV) was considered electrical- repolarisation phase (phase 3) caused by an after-potential. ly homogeneous and isotropic with an intracellular con- A value of H-ratio equal to 0% indicates the absence of an ductivity of 0.1 S/ m and an extracellular conductivity of after-potential, so that the action potential repolarises 0.2 S/ m. smoothly to its resting value. When the after-potential A single action potential template of a myocardial cell causes the slope just before the end of the action potential transmembrane action potential (AP) was associated with phase 3 to become horizontal, the H-ratio measures 100%. each point of the left ventricle. To model repolarisation, In this case the action potential will have a flat hump just the left ventricle was set initially to be completely depolar- before returning to its resting values. A value of the ised with all the action potentials in the plateau phase H-ratio over 100% is possible and denotes a peaked hump (phase 2). A specific repolarisation starting time was set (positive slope), associated with a net inward transmemfor each point on the left ventricular surface (both epi- and brane current. Small values of the H-ratio indicate an endocardium) according to published experimental repo- action potential exhibiting a small after-potential, while larisation sequences [15,16]. large values of the H-ratio denote an action potential The global dispersion of repolarisation times in the exhibiting a large after-potential. Examples of action model was varied by increasing the time delay between the potentials exhibiting after-potentials with different H-ratios first and the last region in the left ventricle to repolarise. are shown in Fig. 1. The repolarisation starting times across the left ventricle between these two extreme regions were linearly interpo Simulations lated. We have now studied this model with after-potentials Three sets of simulations of the 12-lead electrocaradded to the action potentials of all cardiac cells, and diograms were obtained as follows:

3 204 D. di Bernardo, A. Murray / Cardiovascular Research 53 (2002) Fig. 1. Action potentials with simulated after-potentials used for the simulation of U-waves. Top panels: after potential components (APCs) for different values of the scaling factor and time shift. Middle panels: normal action potential (AP). Bottom panels: simulated APs exhibiting after-potentials obtained as the sum of the APCs with the normal AP. Last panel to the right shows a simulated AP exhibiting a delayed after-depolarisation (DAD) obtained with an APC shifted in time by 1100 ms. 1. Changes in after-potential amplitude. The action po Variation of action potential shape tentials used exhibited after-potentials with values of H-ratio varying from 0% (no after-potential) to 98% The action potential shape was varied by separately (almost flat hump). This was achieved by varying the changing the shape of its two components: normal action APC scaling factor from 0 to 1, with a 0.1-step. potential and APC. We modified the shape of these Dispersion of repolarisation was kept constant and components by increasing and decreasing the exponential equal to 20 ms, and no changes to the APC timing were function parameters by 10%. Two action potentials with made. after-potentials were obtained by adding the normal action 2. Changes in dispersion of repolarisation. We changed the potential to the modified APCs and two others by adding value of maximum dispersion of repolarisation in the the modified action potentials to the normal APC. To test left ventricle from 20 to 100 ms in 20-ms steps. The whether the simulated U-waves were sensitive to changes H-ratio was kept constant at 98% (APC scaling factor in shape of the action potential with after-potential, we of 1), and no changes to the APC timing were made. repeated the simulations described above, this time using This simulation was repeated for the action potentials each of the four modified action potentials. This allowed exhibiting after-potentials with values of the H-ratio the computation of the mean and standard deviation (S.D.) less than 98%. The APC scaling factor was varied from of the characteristics measured on the U-waves. 0 (H-ratio 0%) to 1.0 (H-ratio 98%) in 0.1-steps. 3. Changes in timing of after-potentials. We shifted the 2.4. Measurement of U-wave APC in time from 2100 to 100 ms with a 20-ms step. For a 2100-ms time shift of the APC, the resulting To quantify the U-waves simulated by our model, the action potential had a notch interrupting its phase 2. characteristics reported by Surawicz [2] were measured: This is shown in Fig. 1, third column. For a 1100-ms U-wave duration (measured from the T-wave end to the time shift, the resulting AP had an after-potential U-wave end); T-wave end to U-wave peak interval; U- beginning at completion of its repolarisation, i.e. this is wave amplitude expressed as a percentage of the preceding equivalent to a delayed after-depolarisation (DAD). T-wave amplitude. The following definitions were used: This is shown in Fig. 1, fourth column. Dispersion was the T-wave end was defined as the nadir between the kept constant and equal to 20 ms, and the APC scaling T-wave and the U-wave; the U-wave end as the return to factor equal to 1.0. the baseline.

4 D. di Bernardo, A. Murray / Cardiovascular Research 53 (2002) Results show the S.D. of these characteristics, with varying model properties. U-wave duration varied from a mean of 185 to 3.1. Effects of after-potential amplitude 228 ms, T-wave end to U-wave peak interval increased from 35 to 88 ms and U-wave amplitude as a percentage of The after-potential H-ratio is a measure of the after- T-wave amplitude increased from 8 to 18%. For a value of potential amplitude. In the left panels of Fig. 2, the effect the H-ratio less than 65%, U-waves could not be differenof increasing the action potential H-ratio on the simulated tiated from the T-waves. U-wave is summarised: in the top panels, the action potential repolarisation phases for three different and 3.2. Effects of dispersion of repolarisation increasing values of the after-potential H-ratio are shown together with the corresponding T- and U-waves simulated The right panels of Fig. 2 show the effect of increasing by the model. The larger the after-potential, the more maximum dispersion of repolarisation on the simulated distinct the U-wave on the simulated ECG leads. Panels A, U-waves. In the top panels, the last AP in the left ventricle B and C show the characteristics measured on the simu- to repolarise is shown in addition to the first, to indicate lated U-waves in the precordial leads for values of after- the maximum dispersion of repolarisation. Simulated T- potential H-ratio varying from 65 to 98%, and for a and U-waves are shown for values of dispersion equal to constant value of dispersion equal to 20 ms. Error bars 20, 60, and 100 ms. Panels D, E and F show the Fig. 2. Effect of after-potential amplitude (H-ratio) and of dispersion of repolarisation on the simulated U-waves. Action potentials and T and U-waves are shown (x-axis length 600 ms). For changing H-ratio, the dispersion was kept constant at 20 ms. For changing dispersion (first and last action potential to repolarise are shown), the after-potential H-ratio was kept constant and equal to 98%, corresponding to an APC scaling factor of 1.0. Error bars show the standard deviation of the parameters computed using the modified action potentials. There are no values for H-ratio below 65%, as the U-wave was not distinct from the T-wave.

5 206 D. di Bernardo, A. Murray / Cardiovascular Research 53 (2002) characteristics measured on the U-waves in the precordial 3.4. Variation of action potential shape leads for values of dispersion ranging from 20 to 100 ms, with a constant value of after-potential H-ratio equal to Error bars obtained by plotting the standard deviation of 98%. Error bars show the S.D. of these characteristics with the characteristics computed from the U-waves simulated varying model properties. U-wave duration and the interval using the reference and four modified action potentials are from the T-wave end to U-wave peak both decreased, shown in Fig. 2. The exact shapes of the two components indicating that the U-wave became closer to the T-wave. used to generate the action potential exhibiting an after- U-wave amplitude as a percentage of T-wave amplitude potential are not critical for the generation of a U-wave. remained fairly constant with increasing dispersion. This behaviour was also present for values of the afterpotential H-ratio less than 98%. For values between 73 and 4. Discussion 98%, U-wave duration and T-wave end to U-wave peak interval decreased by approximately only 14 ms for a U-waves generated by our model, for low values of change in dispersion from 20 to 100 ms. U-wave amplitude dispersion, resemble closely those observed in normal increased only slightly by 1%. For values of the after- subjects. Simulated U-waves have a shorter ascent than potential H-ratio smaller than 73%, increasing the value of descent, in agreement with clinical observations [2], and dispersion to 100 ms caused the U-wave to merge com- the characteristics measured from our simulated U-waves pletely with the preceding T-wave and therefore to dis- are in the range of the ones measured by Surawicz [2] in appear. When no after-potential was present, i.e. H-ratio normal subjects. equal to 0%, the effect of increased dispersion of repolari- The model demonstrates that the after-potential theory is sation was that of changing the shape of the T-waves by compatible with the generation of normal U-waves on the making them more symmetric [14]. No U-waves were, surface ECG. Our results show that an action potential however, generated. (AP) exhibiting an after-potential prolonging its repolarisation phase is sufficient to generate normal U-waves in the 12-lead ECG. This is in agreement with observations 3.3. Effects of the timing of the after-potential obtained during hypokalemia, when prominent after-potentials appear on the AP [12] and prominent U-waves are The effect of shifting in time the after-potential is observed on the surface ECG. Also, many experimental summarised in Fig. 3, which shows the action potentials studies in long QT syndrome patients (LQTS) have shown used for the simulations and the corresponding simulated that after-potentials exist when U-waves are present [19 T- and U-waves. Negative time shifts cause an interruption 22]. of the rising slope of the T-wave when more negative than This work and our previous work on T-wave shape [14] 260 ms. Increasing the value above 240 ms, normal show that increased dispersion of repolarisation is a very U-waves appear. For increasingly positive time shifts, the unlikely mechanism for the generation of U-waves. Inresulting action potentials exhibit a delayed after-depolar- creasing dispersion of repolarisation in the model without isation (DAD), and the nadir between the T and U-waves after-potentials caused T-wave shape to change and to deepens until a clearly differentiated and inverted U-wave become more symmetrical [14], but not a separate deflecappears. Inverted U-waves are most clearly seen with an tion to appear after the T-wave. after-potential shifted by 100 ms and are similar to the In our model, the left ventricle is homogeneous and inverted U-waves observed in hypertensive subjects [18]. isotropic. Therefore, we were able to use the surface Fig. 3. Simulated T- and U-waves for time shifted after-potentials. In all ECG panels lead V2 is shown (x-axis length 600 ms).

6 D. di Bernardo, A. Murray / Cardiovascular Research 53 (2002) source model [23] to compute the body surface electro- 5. Conclusion cardiogram from a description of the cardiac electric sources on the heart surface only (both epi- and endo- The introduction of dispersion of repolarisation, by cardium), while sources in the heart volume (transmural delaying repolarisation in different regions of the heart region) did not contribute directly to the surface electro- cannot explain the U-wave, and if after-potentials are cardiogram. It is possible that localised repolarisation introduced, changes in dispersion change the U-wave delay in the transmural region could result in deflections characteristics by small amounts in comparison with appearing after the T-wave, if inhomogeneity and aniso- changes in the after-depolarisation itself. However, the tropy of the myocardium were included in the model. presence of after-potentials on the cardiac action potential Nesterenko and Antzelevitch [24] simulated such waves do explain the U-wave polarity and other characteristic using a computer model of a 131-cm surface representing U-wave features. Also, abnormal after-potential timing the transmural wall, with an inhomogeneous varying corresponds with abnormal U-wave inversion. After-poconductivity at the boundary of the M cells region. tentials have been previously linked to U-waves in pa- Although they did not give detailed results, their simulated tients, but not in healthy subjects. Moreover, until now, no U-waves [5] appeared asymmetrical with a longer ascent experimental study has been designed to test the afterthan descent, similar to a small version of the T-wave [2]. potential theory of U-wave genesis. This work provides, This is not compatible with clinical U-waves. The simu- we believe for the first time, a computer model which lated U-waves in our study have shapes similar to those shows that after-potentials can generate normal U-waves published for clinical U-waves, and these are a direct on the surface ECG, and therefore they may be responsible consequence of the presence of after-potentials on the for U-waves also in normal subjects. This model was able action potential. to simulate U-waves with clinically normal characteristics We were able to generate inverted U-waves and to and also abnormal inverted U-waves. Our model suggests attribute their origin to delayed after-depolarisations. To a realistic and simple solution to this century old problem. our knowledge, this is the first time an explanation for U-wave inversion has been proposed. Experimental studies [10] link prolonged ventricular mechanical diastole to Acknowledgements inverted U-waves. One could speculate that this prolonged stretching causes a delay in the activation of stretch D.d.B. was supported by a post-graduate Marie Curie activated channels, thus causing the after-potential to Fellowship granted by the European Commission (C.N. become a DAD, which, as we show, causes inverted ERB4001GT971847). U-waves Limitations References The model of ventricular repolarisation used in this [1] Einthoven W. Die galvanometrische Registrierung des menschlichen work is an oversimplification of the real heart. Inhomo- Electrokardiogram. Pfluger s Arch 1903;99: geneity and anisotropy of the myocardium have not been [2] Surawicz B. U-wave: facts, hypotheses, misconceptions, and mis- modelled and only the left ventricle was simulated. It has nomers. J Cardiovasc Electrophysiol 1998;9: [3] Kataoka H, Yano S. Associations between different status of been shown [25], however, that for the purpose of simulat- myocardial ischemia and ischemia-related negative or positive ing the body surface electrocardiogram these simplifica- epicardial U-waves. J Electrocardiol 1999;32: tions are acceptable. [4] Lepeschkin E. In: Schlant RC, Hurst JW, editors, Advances in The sequence of left ventricular repolarisation used in electrocardiology, New York: Grune and Stratton, 1972, pp. 431 the model is a simplification based on the work of Cowan 448. [5] Antzelevitch C, Sicouri S. Clinical relevance of cardiac arrhythmias et al. [15] where only the epicardium of the left ventricular generated by after-depolarisations: role of M cells in the generation free wall during repolarisation was mapped. The starting of U-waves, triggered activity and torsade de pointes. J Am Coll time of repolarisation on the endocardium is considered to Cardiol 1994;23: be constant, but delayed from the epicardial repolarisation [6] Lazzara R. The U-wave and the M cell. J Am Coll Cardiol as indicated by other researchers [16]. This allowed the 1995;26: [7] Anyukhovsky EP, Sosunov EA, Rosen MR. Regional differences in endocardium-to-epicardium repolarisation gradient to be electrophysiological properties of epicardium, midmyocardium, and simulated effectively. endocardium in vitro and in vivo correlations. Circulation The action potential used was identical over all the 1996;94: endocardial and epicardial surface. This is not likely to [8] Taggart P, Sutton PMT, Opthof T et al. Transmural repolarisation in occur in a real heart [7,20]. Yet we have shown [14] that the left ventricle in humans during normoxia and ischemia. Cardiov- asc Res 2001;50: changes in action potential shape on the endocardial [9] Yan G, Shimizu W, Antzelevitch C. Characteristics and distribution surface do not alter the simulated electrocardiogram ap- of M cells in arterially perfused canine left ventricular wedge preciably. preparations. Circulation 1998;98:

7 208 D. di Bernardo, A. Murray / Cardiovascular Research 53 (2002) [10] Choo MH, Gibson DG. U-waves in ventricular hypertrophy: pos- [18] Miwa K, Miyagi Y, Fujita M et al. Transient terminal U-wave sible demonstration of mechano-electrical feedback. Br Heart J inversion as a more specific marker for myocardial ischemia. Am 1986;55: Heart J 1993;125: [11] Franz M, Burkhoff D, Yue DT et al. Mechanically induced action [19] Habbab MA, El-Sherif N. TU alternans, long QTU, and torsade de potential changes and arrhythmia in isolated and in situ canine pointes: clinical and experimental observations. PACE 1992;15:916 hearts. Cardiovasc Res 1989;23: [12] Coraboeuf E, Deroueaix E, Coulombe A. Acidosis-induced abnor- [20] Shimizu W, Ohe T, Kurita T et al. Epinephrine-induced ventricular mal repolarisation and repetitive activity in isolated dog Purkinje premature complexes due to early after-depolarisations and effects fibres. J Physiol 1980;76: of verapamil and propranolol in a patient with congenital LQTS. J [13] di Bernardo D, Murray A. A computer model for the study of Cardiovasc Electrophysiol 1994;5: cardiac repolarisation. J Cardiovasc Electrophysiol 2000;11:895 [21] Miwa D, Inoue T, Yokoyama M. Monophasic action potentials in 899. patients with torsade de pointes. Jpn Circ J 1994;58: [14] di Bernardo D, Murray A. Explaining the T-wave shape in the ECG. [22] El-Sherif N, Zweiler RH, Craelius W, Gough BW, Henkin R. QTU Nature 2000;403:40. prolongation and polymorphic ventricular tachyarrhythmias due to [15] Cowan JC, Hilton C, Griffiths C et al. Sequence of epicardial bradycardia-dependent early afterdepolarisations. Circ Res repolarisation and configuration of the T wave. Br Heart J 1988;63: ;60: [23] Simms H, Geselowitz D. Computation of heart surface potentials [16] Franz MR, Bargheer K, Rafflenbeul W et al. Monophasic action using the surface source model. J Cardiovasc Electrophysiol potential mapping in human subjects with normal electrocar- 1995;6: diograms: direct evidence for the genesis of the T wave. Circulation [24] Nesterenko V, Antzelevitch C. M-cells as the basis for the electro- 1987;75: cardiographic U-wave. Circulation 1992;86:I 302. [17] Wohlfart B. A simple model for demonstration of ST-T changes in [25] Gulrajani RM. Bioelectricity and biomagnetism, New York: Wiley, the ECG. Eur Heart J 1987;8: