Drug-induced torsades de pointes and implications for drug development

Abstract

Torsades de pointes is a potentially lethal arrhythmia that occasionally appears as an adverse effect of pharmacotherapy. Recently developed understanding of the underlying electrophysiology allows better estimation of the drug-induced risks and explains the failures of older approaches through the surface ECG. This article expresses a consensus reached by an independent academic task force on the physiologic understanding of drug-induced repolarization changes, their preclinical and clinical evaluation, and the risk-to-benefit interpretation of drug-induced torsades de pointes. The consensus of the task force includes suggestions on how to evaluate the risk of torsades within drug development programs. Individual sections of the text discuss the techniques and limitations of methods directed at drug-related ion channel phenomena, investigations aimed at action potentials changes, preclinical studies of phenomena seen only in the whole (or nearly whole) heart, and interpretation of human ECGs obtained in clinical studies. The final section of the text discusses drug-induced torsades within the larger evaluation of drug-related risks and benefits.

Figure 1

Examples of electrocardiograms of torsade de pointes.The top panel shows an episode of the tachycardia on dofetilide, the bottom panel on sotalol. In the bottom panel, torsade starts as a few beats, followed by a pause, followed more beats of the tachycardia, followed by another pause that leads to a prolonged episode of torsade.

Figure 2

Electrical heterogeneity in the canine ventricle: The figure shows computer simulations of the three predominant ventricular cell type: epicardial, M, and endocardial cells. The action-potential distinctions are based on differences in ion-channel activity recorded from the three cell types in voltage-clamp experiments conducted in enzymatically-dissociated canine ventricular myocytes: Ito = transient outward K⁺ current; ICa,L = L-type Ca²⁺ current; Late INa = late sodium channel current; INa-Ca = sodium-calcium exchange current; IK_r = rapidly-activating delayed-rectifier K+ current; IKs = slowly-activating delayed-rectifier K⁺ current; ICl(Ca) = calcium-activated Cl⁻ current; IK1 = inward rectifier K⁺ current; INa-K = electrogenic sodium-potassium exchange (“sodium pump”) current. (Nesterenko and Antzelevitch, unpublished observation).

Figure 3

Sotalol-induced early afterdepolarization in an action potential recorded from a guinea-pig M cell preparation.

Figure 4

Transmembrane potentials and a transmural ECG recorded from an arterially-perfused canine left ventricular wedge preparation. Action potentials from epicardial (Epi), M, and endocardial (Endo) sites were simultaneously recorded using floating glass microelectrodes, together with a transmural ECG. Repolarization of epicardium is coincident with the peak of the T wave of the ECG, whereas repolarization of the M cells coincides with the end of the T wave. Opposing voltage gradients on either side of the M-cell region are responsible for the inscription of the T wave. Reproduced from (135), with permission.

Figure 5

Dose-dependent effect of sodium pentobarbital (10, 20, 50 μg/ml) on transmembrane and ECG activity in an arterially-perfused canine left ventricular wedge preparation. All traces depict action potentials simultaneously recorded from endocardial (Endo), M, and epicardial (Epi) sites together with a transmural ECG. BCL = 2000 msec. Pentobarbital produced a dose-dependent prolongation of the QT interval and APD of all three cell types, but it prolonged the APD of the endocardial and epicardial cells more than that of the M cells, thus reducing transmural dispersion of repolarization (and flattening the T wave) in a dose-dependent manner. Numbers associated with each action potential indicate the APD90 value. Numbers associated with the ECG denote the QT interval, and those beneath the ECG represent the transmural dispersion of repolarization. Reproduced from (6), with permission.

Figure 6

Schematic of arterially-perfused canine left ventricular wedge preparation. Reproduced from (136), with permission

Figure 7

Example of QT/RR hysteresis. The top part of the figure shows leads V4 and V6 of a 10-second electrocardiogram recorded while heart rate was decelerating (consecutive RR intervals of 768, 764, 858, 982, 984, 904, 1144, 1120, and 1056 ms). The bottom part of the figure shows superimposition of the QRS-T complexes following the shortest (764 ms) and longest (1144 ms) RR intervals. Despite the marked differences in RR intervals, the QT interval is constant at 376 ms. Data like these demonstrate the inappropriateness of blindly correcting the QT interval on the basis of the immediately preceding RR interval, or of blindly utilizing all collected intervals in an application of the bin method. Using the Bazett correction, QTc in the tracing shown ranges from 352 ms to 430 ms.

Figure 8

Difference between QT interval in lead V2 and lead II as a function of QT interval in lead II, from manual measurements made on computer screen according to previously published technology (76) in 10 332 digital 10-s 12-lead electrocardiograms. The substantial differences are largely caused by isoelectric projection of the terminal part of the T wave in one of the two compared leads.

Figure 9

Example of problems with automatic measurements by contemporary electrocardiogram machines. The left and middle panels show scans of leads I and II of two automatically-measured electrocardiograms; the right panel shows superimposition of the electrocardiographic patterns from the two tracings. The QT interval duration is the same in both tracings, but the machine reports a difference of 100 ms. Reproduced from Camm AJ, Yap YG, Malik M. Acquired long QT syndrome. Blackwell, 2004, with permission.

Figure 10

Misleading “correction” by Bazett formula in a study of QT/QTc changes induced by atropine. See the text for details.

Figure 11

Heart rate and QT interval measurements in 100 electrocardiograms from each of two normal middle-aged male subjects. Only stable electrocardiograms free of QT/RR hysteresis and of recording noise were selected from day-time continuous 12-lead recordings. Measurement was performed manually using previously described technology.(76)

Figure 12

Regression representation of QT/heart rate relationships in 6 normal middle-aged subjects. The data of each subject were fitted with a regression line corresponding to the individualized form of (Eq 1).