Despite impressive recent advances in antiarrhythmic

Current Perspective Chaos and the Transition to Ventricular Fibrillation A New Approach to Antiarrhythmic Drug Evaluation James N. Weiss, MD; Alan Garfinkel, PhD; Hrayr S. Karagueuzian, PhD; Zhilin Qu, PhD; Peng-Sheng Chen, MD Abstract Sudden cardiac death resulting from ventricular fibrillation can be separated into 2 components: initiation of tachycardia and degeneration of tachycardia to fibrillation. Clinical drug studies such as CAST and SWORD demonstrated that focusing exclusively on the first component is inadequate as a therapeutic modality. The hope for developing effective pharmacological therapy rests on a comprehensive understanding of the second component, the transition from tachycardia to fibrillation. We summarize evidence that the transition from tachycardia to fibrillation is a transition to spatiotemporal chaos, with similarities to the quasiperiodic transition to chaos seen in fluid turbulence. In this scenario, chaos results from the interaction of multiple causally independent oscillatory motions. Simulations in 2-dimensional cardiac tissue suggest that the destabilizing oscillatory motions during spiral-wave reentry arise from restitution properties of action potential duration and conduction velocity. The process of spiral-wave breakup in simulated cardiac tissue predicts remarkably well the sequence by which tachycardia degenerates to fibrillation in real cardiac tissue. Modifying action potential duration and conduction velocity restitution characteristics can prevent spiral-wave breakup in simulated cardiac tissue, suggesting that drugs with similar effects in real cardiac tissue may have antifibrillatory efficacy (the Restitution Hypothesis). If valid for the real heart, the Restitution Hypothesis will support a new paradigm for antiarrhythmic drug classification, incorporating an antifibrillatory profile based on effects on cardiac restitution and the traditional antitachycardia profile (classes 1 through 4). (Circulation. 1999;99:2819-2826.) Key Words: arrhythmia death, sudden drugs dynamics fibrillation Despite impressive recent advances in antiarrhythmic therapies, ventricular fibrillation (VF) remains a major health problem, accounting for 300 000 sudden cardiac deaths annually in the United States. From a conceptual standpoint, treatment of VF has advanced little since the development of electrical defibrillation in the 1950s. Although implantable cardioverter-defibrillators now save many lives, it would be highly desirable and cost-effective to develop less invasive technologies. Pharmacological therapy represents one such avenue but has been largely stalemated by disappointing results of large-scale clinical trials such as CAST 1 and SWORD, 2 which showed increased mortality from sudden death in post myocardial infarction patients treated with class 1 or 3 antiarrhythmic drugs. Why Has the Drug Therapy of Cardiac Fibrillation Been So Disappointing? In pioneering computer modeling studies, Moe and colleagues 3 demonstrated that simulated cardiac tissue could support multiple reentrant wave fronts meandering in complex patterns resembling fibrillation, and they proposed the multiple wavelet hypothesis of fibrillation. This hypothesis was later elegantly validated experimentally by Allessie et al. 4 If fibrillation is due to multiple reentrant wave fronts, how do they arise? Clinical observations indicate that VF is almost always preceded by ventricular tachycardia (VT), the duration of which may vary from a few to many beats 5 (Figure 1). Cineangiographic studies by Wiggers 6 documented that the earliest tachysystolic stage was characterized by rapid coordinated contractions, which later studies 7,8 showed corresponded to a single or double (figure-8) reentrant wave front, subsequently breaking up into multiple fractionated wave fronts characteristic of fully developed fibrillation. These observations suggest that sudden death from VF is separable into 2 components: a triggering event that initiates VT and degeneration of VT to VF (Figure 1). CAST and SWORD 1,2 were therapeutically aimed at suppressing VT initiation. Their failure to prevent sudden cardiac death demonstrates that exclusively targeting VT initiation is not, in general, effective. Why not? In CAST, a positive response to drug therapy was defined as 80% suppression of ventricular ectopy. However, eliminating 80% of events potentially triggering VT translates to a commensurate reduction in VF only if the drug does not adversely potentiate the likelihood that tachycardia, once initiated, will degenerate to fibrillation. Obviously, if the probability of the latter is concomitantly From the Departments of Medicine (Cardiology), Physiology, and Physiological Science and the UCLA Cardiovascular Research Laboratory, UCLA School of Medicine and Cedars-Sinai Medical Center, Los Angeles, Calif. Correspondence to James N. Weiss, MD, Division of Cardiology, 3645 MRL Bldg, UCLA School of Medicine, Los Angeles, CA 90095-1760. E-mail jweiss@mednet.ucla.edu 1999 American Heart Association, Inc. Circulation is available at http://www.circulationaha.org 2819

2820 Chaos and Fibrillation Figure 1. Surface ECG illustrating evolution of human VF from sinus rhythm (SR) to VT to VF. increased by a factor of 5, then the 80% reduction in triggering events is negated. At the present time, little is known about how antiarrhythmic drugs affect the propensity of VT to degenerate into VF. However, an effective pharmacological approach must take antifibrillatory properties of antiarrhythmic drugs, as well as traditional antitachycardic (ie, suppression of VT initiation) properties, into account. This is the area into which nonlinear dynamics and chaos theory is providing new insights. VF: Relationship to Functional Reentry and Spiral Waves The mechanism by which the initial tachysystolic phase degenerates into fibrillation is still poorly understood. In the 1970s, Allessie et al 9 made the seminal observation that functional reentry can occur in cardiac tissue in the absence of an anatomical obstacle, originally called leading circle reentry. This suggested a plausible explanation for the initial tachysystolic phase of fibrillation and raised the possibility that unstable functional reentry breaking up into multiple wave fronts could cause fibrillation. Subsequently, functional reentry was documented experimentally in animal 10 12 and human ventricle. 13 Meanwhile, computer simulations predicted that spiral waves might be the underlying mechanism. 14 Spiral waves are a generic property of excitable media; they arise in cardiac tissue because conduction velocity depends on wave-front curvature. 15 The more convexly curved a wave front is, the more slowly it propagates, because the depolarizing current at the leading edge is diluted into a larger sink of resting cells. At a critical curvature, the source of depolarizing current is too small to bring the sink (resting tissue) to threshold, and propagation fails. A spiral wave is produced when a break occurs along a propagating wave front; at the break, the end is highly curved and forms the tip of the spiral wave, which precesses around a core defined by the critical curvature. Along the arm of the spiral wave, curvature progressively decreases, allowing conduction velocity to increase with radial distance from the tip. The landmark study from Davidenko et al 16 demonstrated spiral-wave reentry as a mechanism of functional reentry in real cardiac tissue. That study also documented that spiral-wave reentry could be unstable, as predicted by computer simulations. 17,18 In this case, the core around which the spiral-wave arm rotates is not stationary but meanders through the tissue, producing the ECG appearance of polymorphic tachycardia or even fibrillation. 19 In addition to meandering, computer simulations also demonstrated that spiral waves can break up to form multiple spiral waves, 17,19 21 similar to wave fronts observed during fibrillation. Movement of multiple spiral waves and their cores (also called phase singularities) is complex, 22 so the classic spiral morphology becomes distorted, characteristic of disorganized wave fronts in fully developed fibrillation (Figure 2). How Do Nonlinear Dynamics and Chaos Theory Apply to Fibrillation? The demonstration that a deterministic computer model, with no random elements, mimics fibrillation raises a strong possibility that fibrillation is not random reentry but arises from a deterministic process, ie, is an example of chaos. We investi- Figure 2. Activation wave fronts (red lines) during fibrillation in canine right ventricle (A) and simulated 2D cardiac tissue (B). Modified from the cover figure of J Clin Invest. 1997;99(2).

Weiss et al June 1, 1999 2821 Figure 3. Deterministic chaos in logistic equation. A, Sequential data set (x n ) with very irregular pattern. B, Poincaré plot of data in 3A in which the current value of x (x n ) is plotted against its previous value (x n 1 ), showing nonlinear parabolic relationship described by equation (inset). C through F, Perioddoubling route to chaos. Time series of x n vs n for different values of, showing transition from period 1 (C) to higher periods (D and E) to chaos (F) as increases. F, Demonstration of sensitivity to initial conditions. Changing initial value (x 0 ) from 0.2000000 to 0.2000001 caused completely different evolution after 20 iterations (thick gray line). Modified from J Clin Invest. 1994;93:1355 1360 (Reference 46) by copyright permission of the American Society for Clinical Investigation. gated this possibility by applying nonlinear dynamics and chaos theory to experimental mapping of cardiac fibrillation and computer simulations. 23 The findings strongly support the idea that cardiac fibrillation represents spatiotemporal chaos. Moreover, the analysis suggests that chaos develops by a specific route, the quasiperiodic transition to chaos. 23,24 The identification of this route to chaos provides very specific predictions about how antiarrhythmic drugs may be designed to prevent cardiac fibrillation. What Is Chaos Theory? Chaos theory is the study of how highly complex behavior, with seemingly random properties, arises from determinism (Figure 3). Determinism means that the current state of a system is determined by its previous state. A simple way to look for a deterministic relationship is to construct a Poincaré plot, in which the present state (x n ) is plotted against the previous state (x n 1 ) (Figure 3B). If the relationship between x n and x n 1 is random, the Poincaré plot yields a formless cloud of points. If the relationship is deterministic, the Poincaré plot may exhibit structure reflecting the underlying dependence of the present state on previous state(s). The unpredictable behavior of a chaotic system results from sensitivity to initial conditions; ie, small perturbations get magnified exponentially (Figure 3F). From a therapeutic standpoint, a key point is that chaos refers to complex behavior that a system may exhibit depending on values of critical system parameters ( in Figure 3). If these critical parameters controlling behavior can be identified and manipulated, chaotic behavior might be eliminated. Another important aspect illustrated in Figure 3 is that the transition from periodic behavior to chaos is usually not abrupt but occurs through an identifiable transition (or bifurcation sequence). A rather astounding fact is that for all known deterministic equations exhibiting chaotic behavior, only a small number of fundamental routes to chaos have been identified. 25 Therefore, identifying the route to chaos may provide important insights into etiology to guide the search for critical parameters governing system behavior. Evidence That Fibrillation Is Chaotic Previous studies addressing the question of whether fibrillation is chaotic have yielded mixed results. 26 29 We looked for chaos by analyzing local extracellular electrograms obtained at single

2822 Chaos and Fibrillation Figure 4. A through D, Sample local electrograms (top) and Poincaré plots of interactivation intervals (bottom) during human atrial fibrillation (A), VF in cross-perfused Langendorff canine heart (B) and canine right ventricular sheet (C), and VF after spiral-wave breakup in simulated 2D cardiac tissue (D). E, Poincaré section constructed from flow velocity measurements during weak fluid turbulence (modified from Reference 47, with permission from the American Physical Society). F, Origin of the ringlike pattern in a Poincaré section of a quasiperiodic system. If the first oscillatory motion is represented by circular trajectory of vector (arrow) in horizontal plane and the second by a circular trajectory in the vertical plane, their combined motion defines a circle of circles, or torus. The trajectory of the measured variable (eg, voltage) lives on surface of the torus. For a system sampled periodically at defined events such as a QRS deflection (indicated by the plane transecting the torus), then with each rotation around the torus, the trajectory pierces the sampling plane, producing a ring with hole in center. When a third oscillatory motion is added, the torus thickens, develops wrinkles, or is destroyed, and chaos ensues. 25 Modified from J Clin Invest. 1997;99:305 314 (Reference 23) by copyright permission of the American Society for Clinical Investigation. recording sites in 3 types of hemodynamically stable fibrillation: human atrial fibrillation, canine VF with maintained coronary blood flow, and VF in ventricular muscle sheets. 23 The first step was to construct Poincaré plots of the interactivation intervals measured from the recorded electrograms. As shown by Figure 4A through 4C, in all 3 types of fibrillation, the Poincaré plots did not show a diffuse cloud of points suggesting a random process but instead had definite structure: a ring of points with a hole in the center. 23 Computer simulations showed that the ringlike pattern was not an artifact. 23 When 2-dimensional (2D) cardiac tissue was simulated with a ventricular action potential model, spiralwave reentry was readily initiated and broke up into multiple reentrant wave fronts resembling fibrillation (Figure 2). When interactivation intervals at a local site were analyzed, the Poincaré plot also had a ringlike appearance (Figure 4D). The Route to Chaos in Fibrillation The ringlike structure in the Poincaré plot has particular significance, because it suggests that chaos arose by a quasiperiodic transition. This route to chaos was first described in a theoretical treatment of fluid turbulence by Newhouse et al. 30 They proved that a system containing 3 independent oscillatory modes is inherently unstable and becomes chaotic. Figure 5A illustrates that theorem with a stream of smoke injected into air. Near the origin of the stream, the flow pattern is laminar, but a bit farther away, a pair of stationary vortices form (first oscillatory motion). A bit farther still, the vortex develops a waviness (second oscillatory motion), which grows in amplitude and complexity (third oscillatory motion), upon which a sudden transition to full turbulence occurs. An analogous process was observed in simulated fibrillation. 23 Figure 5B shows the initiation of a spiral wave (first oscillatory motion). The thickness of the spiral-wave arm represents the amount of depolarized tissue between the action potential upstroke and repolarization phase, also known as the wavelength (the product of action potential duration [APD] and conduction velocity [CV]). Along the spiral-wave arm, wavelength is not uniform but is thicker in some regions and thinner in others, as in the smoke vortices. As a result, the wave front

Weiss et al June 1, 1999 2823 Figure 5. A, Development of turbulence in smoke stream, illustrating a quasiperiodic transition to chaos. Sequence from left to right shows laminar flow (L), stable vortex (first oscillation, O1), waviness in vortex walls (second oscillation, O2), even more complex oscillatory motion of vortex wall (third oscillation, O3), and abrupt transition to full turbulence (T). Modified from Reference 48, by permission of Dr H. Nagib. B through E, Quasiperiodic transition to chaos during spiral-wave breakup in 2D simulated cardiac tissue (300 300 cells). Four snapshots are at times indicated after initiation of spiral wave. Note increasing oscillations in wavelength (white-gray area) along spiral-wave arm (B and C), eventually leading to wave break and formation of daughter spiral waves (C through E). and the wave back of the spiral arm develop a scalloped appearance (secondary and tertiary oscillatory motions), which are progressively amplified until at 1 point along the spiral arm, the wavelength becomes too short to propagate (Figure 5C). At this break in the spiral arm, the 2 ends form the tips of 2 daughter spiral waves (Figure 5D). This breakup process repeats itself until fibrillation is fully developed (Figure 5E). Figure 6 shows an analogous process in a local electrogram recorded during a spontaneous transition from VT to VF in human right ventricular myocardium. 23 During VT, the electrogram is monomorphic with a constant cycle length (CL) (first oscillatory motion). At the first arrow, a transient irregularity developed that settled into a period-2 bigeminal rhythm (second oscillatory motion). This period-2 rhythm then developed low-frequency amplitude and period modulation (third oscillatory motion), which grew progressively larger until the electrogram suddenly became completely irregular, abruptly marking the onset of VF. In fluid turbulence, as in cardiac fibrillation, the Poincaré plot constructed from flow measurements typically shows a ringlike structure (Figure 4E). More rigorous mathematical criteria, such as nonmonotonic circle maps, 25 also supported the hypothesis that the route to chaos in fibrillation arises through quasiperiodicity. 23 What Are the Oscillatory Motions Producing Chaos in Cardiac Fibrillation? The close agreement between simulation and experiment supports spiral-wave breakup in simulated cardiac tissue as a reasonable model for the transition from tachycardia to fibrillation. We hypothesize that the process by which a spiral wave (the tachysystolic phase of fibrillation) meanders and then breaks up into multiple reentrant wave fronts (fully developed fibrillation) represents a quasiperiodic transition to spatiotemporal chaos, resulting from the interaction of multiple coupled oscillatory motions. Understanding the origin of these oscillations could, in principle, provide the key to preventing fibrillation. As illustrated in Figure 5B through 5E, the additional oscillatory motions that cause spiral-wave breakup are directly observed as oscillations in wavelength along spiral-wave arm soon after its initiation. This is an important clue to their cellular basis. Because wavelength is by definition the product of APD and CV, then for wavelength to change requires 1 or both parameters to change. APD and CV are each determined by their restitution properties. Therefore, APD and CV restitution must be the source of wavelength oscillations that lead to spiral-wave breakup. One method for measuring APD restitution is illustrated in Figure 7A, in which a premature stimulus (S2) is delivered at progressively shorter diastolic intervals (DIs) during pacing at a fixed CL. For short DIs (defined as the interval between the end of the previous action potential and S2), the action potential does not recover its full amplitude or duration. The plot of APD versus DI is an APD restitution curve (Figure 7B). In cardiac tissue, the slower velocity of the action potential upstroke at short DIs also slows CV, and a similar curve of CV versus DI defines the CV restitution curve (Figure 7C). It is known that the steepness of APD restitution has played a critical role in stability of reentry around an anatomic obstacle in both experimental 31,32 and theoretical studies. 33,34 Steep APD restitution (slope 1) promotes instability, which can terminate reentry around an obstacle (Figure 8). Karma 20 realized that the same instability, if it occurred along the arm of a spiral wave, would produce wave break, the essential event in spiral-wave breakup. To understand the mechanism, we can think of APD restitution as a difference amplifier, providing the next value of APD as a function of the previous DI. For a spiral wave rotating at a constant CL, the equilibrium values of APD and DI occur at the intersection of the APD restitution curve, with the dashed line representing CL (Figure 8). Suppose there is a slight perturbation in DI. Because of APD restitution, the new DI will cause the APD of the next beat to differ. The new APD will then generate a new DI. Whether this difference in DI is greater or smaller is determined by the APD restitution

2824 Chaos and Fibrillation Figure 6. Quasiperiodic transition to chaos during degeneration of VT to VF in human right ventricular tissue. Local extracellular electrogram tracing was on the ventricular surface. After initiation of VT (first oscillation), a spontaneous transient instability (bar above upper trace) settled into an alternating period-2 pattern (second oscillation). A third lowfrequency amplitude and period modulation then developed (third oscillation), which grew progressively larger (bar above lower trace), presaging abrupt transition to fibrillation (arrow). Reproduced from J Clin Invest. 1997;99:305 314 (Reference 23) by copyright permission of the American Society for Clinical Investigation. slope. If it is 1 (Figure 8A), the next difference is smaller, and if it is 1 (Figure 8B), the next difference is larger. In this way, a shallow slope restores APD and DI back to their equilibrium values, whereas a steeply sloped APD restitution curve amplifies the differences so that they progressively diverge (ie, the equilibrium is unstable). If this oscillation grows large enough, the DI will eventually become shorter than the refractory period, causing a wave break at some point along the spiral-wave arm. APD restitution produces oscillations in the wave back (repolarization phase), whereas CV restitution creates oscillations in the wave front (depolarization phase) by slowing CV in regions with short DIs. This creates a spatial mode of CL oscillation. The interaction between APD and CV restitution creates spatiotemporal oscillations, which are quasiperiodic. A prediction of this hypothesis is that altering cardiac restitution properties should alter the behavior of spiral waves. 20,35 Figure 7 shows that decreasing the Ca current by 50% in the action potential model 36 reduces the range of DIs over which APD restitution is steep (Figure 7B). This prevents spiral-wave breakup, because the DIs experienced along the spiral-wave arm are no longer in a steep range of APD restitution. Wavelength also plays another important role in spiralwave reentry by setting the minimal space required for a spiral wave to sustain itself. If wavelength is too long relative to tissue size, tachycardia or fibrillation will selfterminate 24,37 the basis of the critical mass hypothesis. 38 In 2D simulations, however, spiral-wave dynamics (stationary, meandering, or breakup phenotypes) is determined primarily by underlying electrophysiological properties such as restitution, regardless of tissue size. Can These Findings Help Us to Develop Effective Antifibrillatory Drugs? Sudden cardiac death caused by VF is separable into 2 components: initiation of tachycardia and degeneration of tachycardia to fibrillation. Clinical antiarrhythmic drug studies focusing exclusively on preventing VT initiation have not been effective at reducing sudden cardiac death, so a better understanding of the second component is critically needed to advance pharmacological therapy. The degeneration of tachycardia to fibrillation can be simulated in computer models by spiral-wave breakup, which occurs by a specific route known as the quasiperiodic transition to chaos. Experimental evidence directly supports this scenario. In simulations, APD and CV restitution creates spatiotemporal oscillations in APD and DI, leading to wave break and the onset of a fibrillationlike state. That is, they are potential critical parameters that govern the transition between periodicity (tachycardia) and chaos (fibrillation). The key question is as follows: Is this scenario relevant to real cardiac tissue? If so, it should be possible to develop antiarrhythmic drugs that modify cardiac restitution properties appropriately to prevent tachycardia from degenerating to fibrillation. A strategy for the ideal antiarrhythmic drug would be to combine traditional antitachycardic properties (classes 1 through 4) with antifibrillatory properties on the basis of their effects on restitution as a 2-pronged attack. Like Karma, 20 we refer to this scenario as the Restitution Hypothesis. What Critical Evidence Is Needed to Validate the Restitution Hypothesis? Methods for measuring restitution properties need fuller evaluation. APD restitution is not a unique function of the previous DI but is significantly influenced by excitation

Weiss et al June 1, 1999 2825 Figure 7. Effects of APD and CV restitution on spiral-wave stability in Luo-Rudy ventricular action potential model. A, During pacing at fixed CL (S1-S1), a premature stimulus (S2) is introduced at progressively shorter DIs. APD progressively shortens as DI decreases until cell is refractory. B, APD restitution curve before (solid line) and after (dashed line) Ca current is reduced by 50%. C, CV restitution curves under identical conditions. D and E, Snapshots of activation patterns 5 seconds after initiation of spiral-wave reentry. Under control conditions (D), spiral wave breaks up into multiple wave fronts resembling fibrillation; after Ca current is reduced to flatten APD restitution (E), the spiral wave remains intact. history (memory effect). The restitution characteristics present during the arrhythmia determine stability or instability, and these characteristics may differ significantly from restitution measured by arbitrarily devised pacing protocols. In our own survey of 25 studies, we found that maximal APD restitution slope was either considerably 1 (15 studies) or 1 (10 studies). In none was the slope substantially 1 at all CLs tested. The major cellular determinants of APD and CV restitution are well characterized and include recovery from inactivation of inward Na and Ca currents, deactivation of K currents, and intracellular Ca dynamics. 36,39,40 With a directed approach, antiarrhythmic drugs that favorably alter factors controlling restitution should be identifiable. The real heart is highly complex (3-dimensional, anisotropic, and electrophysiologically and anatomically heterogeneous) compared with simulated 2D tissue for which the Restitution Hypothesis is demonstrated to be valid. Others argue that fibrillation requires 3 dimensions, 2D spiralwave breakup has limited relevance to fibrillation, 41 or 3 dimensions produce restitution-independent breakup. 42 However, fibrillation has been documented in relatively thin-walled cardiac preparations. 24,43 This issue needs to be resolved. Figure 8. Relationship between steepness of APD restitution and spiral-wave stability. For spiral wave rotating at constant CL, APD and DI equilibrate at intersection of restitution curve (solid line) with dotted line defined by relationship APD DI CL. A, For a shallow APD restitution (slope 1), a small perturbation (a) that shortens DI results in even smaller change in APD, producing smaller change in DI, and so forth, until equilibrium is reestablished. B, For steep APD restitution (slope 1), a small decrease in DI (a) produces larger change in APD, which produces larger change in DI, etc. The oscillation is amplified rather than damped. When DI becomes too short to generate action potential (ie, at b), conduction fails, causing wave break along spiral-wave arm. Electrophysiological and anatomic heterogeneity makes the diseased heart more susceptible to VF and sudden death. 44 Although heterogeneity makes reentry easier to induce, its effects on VT stability are less clear. In healthy human heart, functional reentry can invariably be induced with a sufficiently strong stimulus, and once initiated, it virtually always degenerates to VF. In contrast, stable VT is much more common in diseased hearts despite their considerably greater heterogeneity. Heterogeneity can actually promote stability by anchoring meandering spiral waves. 16,45 Therefore, the common perception that heterogeneity intrinsically destabilizes reentrant wave fronts may be a misconception. The higher incidence of VT and VF in diseased hearts may reflect the greater likelihood that a triggering event (premature ventricular contraction) will initiate a sustained arrhythmia rather than a destabilizing effect of heterogeneity on reentry. Conclusions Therapeutically, the role of electrophysiological and anatomic tissue heterogeneity in degeneration of VT to VF is the key challenge facing the restitution hypothesis. If heterogeneity by itself directly promotes spiral-wave instability, this is a difficult therapeutic target. On the other hand, if modifiable global electrophysiological properties such as APD and CV restitution determine stability, the outlook for effective pharmacological treatment of VF is bright. Resolving this issue is critical. It is testable with simulations incorporating realistic cardiac anatomy and electrophysiology, in combination with experimental arrhythmia mapping experiments. From these findings, we recommend that comprehensive experimental evaluation of the Restitution Hypothesis is warranted and timely. A new paradigm for antiarrhythmic drug classification, incorporating an antifibrillatory profile based on cardiac restitution with the traditional antitachycardia profile (classes 1 through 4), seems highly promising. Potentially, this could resolve the stalemate over

2826 Chaos and Fibrillation antiarrhythmic drug development that has followed in the discouraging wake of the CAST and SWORD trials. 1,2 Acknowledgments This work was supported by NIH SCOR in Sudden Cardiac Death P50 HL-52319, grants from the American Heart Association and its Greater Los Angeles Affiliate, the University of California Tobacco- Related Disease Research Program, Cedars-Sinai Foundation ECHO Award, the Ralph M. Parsons Foundation, the Laubisch Fund, and the Kawata and Price endowments. References 1. Cardiac Arrhythmia Suppression Trial (CAST) Investigators. Effect of encainide and flecainide on mortality in a randomized trial of arrhythmia suppression after myocardial infarction. N Engl J Med. 1989;321: 406 412. 2. Waldo AL, Camm AJ, deruyter H, Friedman PL, Macneil DJ, Pauls JF, Pitt B, Pratt CM, Schwartz PJ, Veltri EP. Effect of d-sotalol on mortality in patients with left ventricular dysfunction after recent and remote myocardial infarction. Lancet. 1996;348:7 12. 3. Moe GK, Rheinboldt WC, Abildskov JA. A computer model of atrial fibrillation. Am Heart J. 1964;67:200 220. 4. Allessie MA, Lammers WJEP, Bonke FIM, Hollen J. Experimental evaluation of Moe s multiple wavelet hypothesis of atrial fibrillation. In: Zipes DP, Jalife J, eds. Cardiac Arrhythmias. New York: Grune & Stratton; 1985:265 276. 5. Kempf FC, Josephson ME. Cardiac arrest recorded on ambulatory electrocardiograms. Am J Cardiol. 1984;53:1577 1582. 6. Wiggers CJ. Studies of ventricular fibrillation caused by electric shock, II: cinematographic and electrocardiographic observation of the natural process in the dog s heart: its inhibition by potassium and the revival of coordinated beats by calcium. Am Heart J. 1930;5:351 365. 7. Chen P-S, Wolf PD, Dixon EG, Danieley ND, Frazier DW, Smith WM, Ideker RE. Mechanism of ventricular vulnerability to single premature stimuli in open-chest dogs. Circ Res. 1988;62:1191 1209. 8. Hwang C, Fan W, Chen P-S. Protective zones and the mechanisms of ventricular defibrillation. Am J Physiol. 1996;271:H1491 H1497. 9. Allessie M, Bonke F, Schopman F. Circus movement in rabbit atrial muscle as a mechanism of tachycardia, II: the role of nonuniform recovery of excitability in the occurrence of unidirectional block as studied with multiple microelectrodes. Circ Res. 1976;39:168 177. 10. Mehra R, Zeiler RH, Gough WB, El-Sherif N. Reentrant ventricular arrhythmias in the late myocardial infarction period. Circulation. 1983; 67:11 24. 11. Dillon SM, Allessie MA, Ursell PC, Wit AL. Influences of anisotropic tissue structure on reentrant circuits in the epicardial border zone of subacute canine infarcts. Circ Res. 1988;63:182 206. 12. Schalij M, Lammers W, Rensma P, Allessie M. Anisotropic conduction and reentry in perfused epicardium of rabbit left ventricle. Am J Physiol. 1992;263:H1466 H1478. 13. Downar E, Harris L, Mickleborough L, Shaikh N, Parson I. Endocardial mapping of ventricular tachycardia in the intact human ventricle: evidence for reentrant mechanisms. J Am Coll Cardiol. 1988;11:783 791. 14. Winfree AT. When Time Breaks Down. Princeton, NJ: Princeton University Press; 1987. 15. Cabo C, Pertsov AM, Baxter WT, Davidenko JM, Gray RA, Jalife J. Wave-front curvature as a cause of slow conduction and block in isolated cardiac muscle. Circ Res. 1994;75:1014 1028. 16. Davidenko JM, Pertsov AV, Salomonsz JR, Baxter W, Jalife J. Stationary and drifting spiral waves of excitation in isolated cardiac muscle. Nature. 1992;355:349 351. 17. Courtemanche M, Winfree AT. Reentrant rotating waves in a Beeler- Reuter based model of two-dimensional cardiac electrical activity. Int J Bifurc Chaos. 1991;1:431 444. 18. Efimov IR, Krinsky VI, Jalife J. Dynamics of rotating vortices in the Beeler-Reuter model of cardiac tissue. Chaos Solitons Fractals. 1995;5: 513 526. 19. Gray RA, Jalife J, Panfilov AV, Baxter WT, Cabo C, Davidenko JM, Pertsov AM. Mechanisms of cardiac fibrillation. Science. 1995;270: 1222 1223. 20. Karma A. Electrical alternans and spiral wave breakup in cardiac tissue. Chaos. 1994;4:461 472. 21. Panfilov AV, Hogeweg P. Scroll breakup in a three-dimensional excitable medium. Physiol Rev E. 1996;53:1740 1743. 22. Gray RA, Pertsov AM, Jalife J. Spatial and temporal organization during cardiac fibrillation. Nature. 1998;392:75 78. 23. Garfinkel A, Chen P-S, Walter DO, Karagueuzian HS, Kogan B, Evans SJ, Karpoukhin M, Hwang C, Uchida T, Gotoh M, Nwasokwa O, Sager P, Weiss JN. Quasiperiodicity and chaos in cardiac fibrillation. J Clin Invest. 1997;99:305 314. 24. Kim Y-H, Garfinkel A, Ikeda T, Wu T-J, Athill CA, Weiss JN, Karagueuzian HS, Chen P-S. Spatiotemporal complexity of ventricular fibrillation revealed by tissue mass reduction in isolated swine ventricle. J Clin Invest. 1997;100:2486 2500. 25. Hilborn RC. Chaos and Nonlinear Dynamics. New York, NY: Oxford University Press; 1994:252 294. 26. Kaplan DT, Cohen RJ. Is fibrillation chaos? Circ Res. 1990;67:886 892. 27. Witkowski FX, Penkoske PA. Activation patterns during ventricular fibrillation. Ann N Y Acad Sci. 1990;591:219 231. 28. Karagueuzian HS, Khan SS, Peters W, Mandel WJ, Diamond GA. Nonhomogeneous local atrial activity during acute atrial fibrillation: spectral and dynamic analysis. Pacing Clin Electrophysiol. 1990;13:1937 1942. 29. Witkowski FX, Kavanagh KM, Penkoske PA, Plonsey R, Spano ML, Ditto WL, Kaplan DT. Evidence for determinism in ventricular fibrillation. Physiol Rev Lett. 1995;75:1230 1233. 30. Newhouse S, Ruelle D, Takens F. Occurrence of strange axiom-a attractors near quasiperiodic flow on T m,m 3. Comm Math Phys. 1978; 64:35 40. 31. Simson MB, Spear JF, Moore EN. Stability of an experimental atrioventricular reentrant tachycardia in dogs. Am J Physiol. 1981;240: H947 H953. 32. Frame LH, Simson MB. Oscillations of conduction, action potential duration, and refractoriness: a mechanism for spontaneous termination of reentrant tachycardias. Circulation. 1988;78:1277 1287. 33. Courtemanche M, Glass L, Keener J. Instabilities of a propagating pulse in a ring of excitable media. Physiol Rev Lett. 1993;70:2182 2185. 34. Qu Z, Weiss JN, Garfinkel A. Spatiotemporal chaos in a simulated ring of cardiac cells. Physiol Rev Lett. 1997;78:1387 1390. 35. Courtemanche M. Complex spiral wave dynamics in a spatially distributed ionic model or cardiac electrical activity. Chaos. 1996;6: 579 600. 36. Luo CH, Rudy Y. A model of the ventricular cardiac action potential: depolarization, repolarization, and their interaction. Circ Res. 1991;68: 1501 1526. 37. Rensma PL, Allessie MA, Lammers WJEP, Bonke FIM, Schalij MJ. The length of the excitation wave as an index for the susceptibility to reentrant atrial arrhythmias. Circ Res. 1988;62:395 410. 38. Garrey W. The nature of fibrillatory contraction of the heart: its relation to tissue mass and form. Am J Physiol. 1914;33:397 414. 39. Luo CH, Rudy Y. A dynamical model of the cardiac ventricular action potential, I: simulations of ionic currents and concentration changes. Circ Res. 1994;74:1071 1096. 40. Saitoh H, Bailey JC, Surawicz B. Action potential duration alternans in dog Purkinje and ventricular muscle fibers: further evidence in support of two different mechanisms. Circulation. 1989;80:1421 1431. 41. Winfree AT. Electrical turbulence in three-dimensional heart muscle. Science. 1994;266:1003 1006. 42. Fenton F, Karma A. Vortex dynamics in three-dimensional continuous myocardium with fiber rotation: filament instability and fibrillation. Chaos. 1998;8:20 47. 43. Ikeda T, Wu TJ, Uchida T, Hough D, Fishbein MC, Mandel WJ, Chen P-S. Unstable reentrant wave fronts induced by acetylcholine in isolated canine right atrium. Am J Physiol. 1997;273:H356 H370. 44. Janse MJ, Wit AL. Electrophysiological mechanisms of ventricular arrhythmias resulting from myocardial ischaemia and infarction. Physiol Rev. 1989;69:1049 1168. 45. Ikeda T, Yashima M, Uchida T, Hough D, Fishbein MC, Mandel WJ, Chen P-S, Karagueuzian HS. Attachment of meandering reentrant wave fronts to anatomic obstacles in the atrium: role of the obstacle size. Circ Res. 1997;81:753 764. 46. Weiss JN, Garfinkel A, Spano ML, Ditto WL. Chaos and chaos control in biology. J Clin Invest. 1994;93:1355 1360. 47. Brandstater A, Swinney HL. Strange attractors in weakly turbulent Couette-Taylor flow. Physiol Rev A. 1987;35:2207 2220. 48. Van Dyke M. An Album of Fluid Motion. Stanford, Calif: Parabolic Press; 1998:60.