Independent derivations of the axis of arrhythmia for predicting drug-induced torsades de pointes

Abstract

Torsades de pointes is a potentially lethal ventricular arrhythmia that may be caused by many classes of drugs. The likelihood of arrhythmia depends upon the particular combination of cardiac ion-channels that the drug targets. The risk for a given drug can be measured against the axis of arrhythmia, which is a conceptual line in the parameter space of ventricular cardiomyocyte electrophysiology. The axis describes the most potent combination of ion-channel blocks that is theoretically possible. It thus serves as a convenient yardstick for quantifying the pro-arrhythmic risk of novel drugs from their channel-blocking profiles. To date, the axis of arrhythmia has only been derived from biophysical computer simulations. Here, we derive the axis directly from the response profiles of selected ion-channels to known drugs (n=109). The new method provides an independent line of evidence for estimating the axis of arrhythmia. Using step-wise reduction, the two methods were found to converge to identical estimates of the axis in two ion-currents (ICaL and IKr). The resulting axis predicted the pro-arrhythmic risk of the test drugs with 89.9% to 91.7% accuracy, depending on the dose at which the drug was assessed. The axis of arrhythmia offers a practical method for predicting the pro-arrhythmic risk of novel drugs without the need for drug-specific computer simulations. It is the only metric of pro-arrhythmic risk that has been derived from both a biophysical model and a purely statistical model. It thus combines the benefits of biophysical interpretation and computational efficiency in assessing the torsadogenic risk of new drugs prior to clinical trials. Significance StatementMany classes of drugs can cause lethal cardiac arrhythmias by blocking ion-channels that are needed by heart cells for normal rhythm. International safety guidelines thus require all new drugs be tested in animal hearts or cell cultures prior to undertaking clinical trials in humans. Computer models offer an ethical alternative for drug testing that is gaining acceptance in the safety pharmacology community. However, the current corpus of models are usually too complex to apply outside of specialist computing laboratories. We recently proposed a new approach that eliminates the need for drug-specific computer simulations by using a large one-off computer simulation to identify the most pro-arrhythmic drug that is theoretically possible. Thereafter, that hypothetical drug can be used to assess the pro-arrhythmic risk of real drugs using only pen-and-paper calculations. In that study, the hypothetical drug was described by the axis of arrhythmia in the cardiac cell model. In the present study, we propose an alternative method for deriving the axis of arrhythmia directly from drug datasets. The two derivations are independent, yet they arrive at the same result. Together, they provide converging lines of evidence in support of this new approach to drug testing in safety pharmacology.