On a Newly Proposed In Vitro Proarrhythmia Model

A growing awareness of the potential for proarrhythmia by noncardiovascular drugs has resulted in relabelings, warnings, and withdrawals of some drugs from the market. At issue is a rare, drug-induced ventricular tachyarrhythmia known as torsade de pointes, which (typically) may lead to syncope, or (rarely) progress to ventricular fibrillation and sudden cardiac death. Torsade de pointes has been linked to delayed cardiac repolarization, manifest as prolongation of the QT interval on the electrocardiogram. As a consequence, drug-induced QT prolongation is generally considered as a surrogate marker for proarrhythmia and torsade de pointes. Recognition of the link between delayed repolarization and proarrhythmia had led to the use of in vitro models to detect repolarization changes to predict (and prevent) proarrhythmia. A recent survey of pharmaceutical industry practice (1), opinion articles related to these issues (2,3), and draft guidelines from regulatory authorities (4,5) discuss the role of three preclinical assays typically used to evaluate proarrhythmic potential of noncardiovascular drugs. These include (a) an assay evaluating block of an outward repolarizing potassium current which determines cardiac repolarization (typically hERG [human ether-a-go-go–related gene] or native iKr); (b) a repolarization assay that evaluates changes in the action potential duration (APD) in an integrated electrophysiologic system (such as Purkinje fibers or guinea pig papillary muscles); and (c) an assay evaluating changes in the QT interval on the electrocardiogram recorded in vivo. The need for three assays highlights the fact that no one assay appears to be totally predictive of the clinical experience. For example, while most drugs that block HERG also delay repolarization and are linked to torsade de pointes, some (such as verapamil and fluoxetine) are not generally recognized as demonstrating significant proarrhythmic potential. Other drugs (such as terfenadine) are prominently linked to proarrhythmia in humans, yet fail to elicit prominent prolongation of the action potential duration of nonhuman cardiac tissues in vitro. In addition, the interpretation and significance of small changes in the QT interval (for example, tens of milliseconds or less from a baseline range of 400 ms in humans) are complicated by changes in heart rate, which by themselves affect the QT interval, necessitating the use of a variety of correction factors not necessarily applicable to the species under study. It is likely that no one model alone is fully predictive of proarrhythmic risk since multiple factors likely converge to provide the conditions necessary to initiate or sustain torsade de pointes for most drugs. In this issue of the Journal of Cardiovascular Pharmacology, Hondeghem and Hoffmann (6) evaluate the electrophysiologic effects of fourteen clinically used drugs in a recently described in vitro proarrhythmia model (7). This model consists of a paced Langendorff-perfused female rabbit heart from which monophasic action potentials are recorded; parameters evaluated include action potential duration, conduction, instability (indicative of APD variability), triangulation (indication of changes in configuration), and reverse use dependence. Fourteen blinded drugs were evaluated, including drugs associated with proarrhythmia at therapeutic or excessive concentrations (including terfenadine and haloperidol) as well as those not associated with proarrhythmia (including penicillin and aspirin). The automated analysis program identified each of the 14 blinded drugs in regards to effects on APD and conduction and, more importantly, distinguished drugs generally associated with proarrhythmia based on effects on instability, triangulation, and reverse use dependence. None of the drugs considered as safe from proarrhythmia were labeled as proarrhythmic. As with any study, some limitations should be recognized. First, supratherapeutic (100-to 1,000-fold) concentrations of some compounds (e.g., terfenadine) were needed to evoke instability and proarrhythmia. Without the hindsight of prior clinical identification of “QT-offending drugs,” it is unlikely that such high concentrations would be routinely tested as part of the evaluation of developing drug candidates. This model uses monophasic action potential recordings (rather than transmembrane recordings using microelectrode techniques) to measure action potential changes. In general, monophasic action potential recordings are more difficult to interpret when repolarization abnormalities are present (e.g., early afterdepolarizations and delayed afterdepolarizations) and do not provide traditional information gleaned from microelectrode studies (changes in such parameters as resting membrane potential and maximum upstroke velocity used to infer changes in inward rectifier and fast sodium currents, respectively). While the reproducibility of the model was evaluated by comparing duplicate sets of experiments, a greater number of studies would provide more confidence. In addition, the likely effects of plasma protein binding (expected to diminish the free drug concentration in vivo) on the electrophysiologic actions of drugs are not evaluated in these studies. Finally, these studies were performed solely using hearts from female rabbits, a species that may be more sensitive to delayed repolarization (8). While this greater sensitivity did not produce any false-positive results for the six “non–QT offenders” studied, further evaluations with additional compounds are necessary to guard against overestimating the extent of proarrhythmic risk. It would be valuable to know how this model behaves with additional recognized risk factors for proarrhythmia present (e.g., hypokalemia). Despite the above limitations, the model brings new insights (and questions) to the in vitro evaluation of proarrhythmic risk. The concept of drug-induced instability represents a dynamic measure of a drug's effect on repolarization typically not considered when evaluating proarrhythmic potential using regular stimulation protocols. Poincarré plots (which plot APD values for each beat versus its prior beat) derived from applied stimulation protocols represent a novel approach for displaying a drug's dynamic electrophysiologic effects, and present a future challenge for quantification and validation. Somewhat disappointing is the minimal effect of heroic concentrations of terfenadine on the Poincarré plot compared with those of less excessive concentrations of haloperidol (Fig. 2 of Hondeghem and Hoffman (6)). Measures of instability may more closely reflect the proarrhythmic effects of cardiac rhythms (short–long–short cycle lengths) recognized as risk factors for the initiation of torsade de pointes. It seems likely that similar measures of instability could be obtained using basic stimulation protocols, which included interpolated beats and standard transmembrane recording techniques applied to syncytial in vitro preparations. Where does (or should) this new model fit within the present paradigm to evaluate proarrhythmic potential? Clearly, the present report provides an additional in vitro proarrhythmia model on an organ level of greater complexity beyond that of ionic current measurements and syncytial repolarization assay. However, it remains to be determined whether this in vitro model provides substantial additional useful information compared with results from in vitro and in vivo assays presently in use when considered in concert. Clearly, the risk of detecting a “false-positive” or irrelevant signal increases with increasing numbers of assays applied. Such signals may inappropriately influence consideration of the potential risk versus therapeutic benefit of advancing drug candidates, leading to the unnecessary “overlabeling” of drugs and subsequent lesser impact of such cautions. The utility (and burden) of an additional fourth assay to detect proarrhythmia must be balanced against the predictability of presently established models through further experimentation by different laboratories with multiple compounds. Only with this additional information will it be possible to determine the most representative and efficient preclinical assays to detect proarrhythmic risk.