Replacing the thorough QT study: reflections of a baby in the bath water

Prediction of arrhythmia susceptibility through mathematical modeling and machine learning

At present, the QT interval on the electrocardiographic (ECG) waveform is the most common metric for assessing an individual's susceptibility to ventricular arrhythmias, with a long QT, or, at the cellular level, a long action potential duration (APD) considered high risk. However, the limitations of this simple approach have long been recognized. Here, we sought to improve prediction of arrhythmia susceptibility by combining mechanistic mathematical modeling with machine learning (ML). Simulations with a model of the ventricular myocyte were performed to develop a large heterogenous population of cardiomyocytes (n = 10,586), and we tested each variant's ability to withstand three arrhythmogenic triggers: 1) block of the rapid delayed rectifier potassium current (I_Kr Block), 2) augmentation of the L-type calcium current (I_CaL Increase), and 3) injection of inward current (Current Injection). Eight ML algorithms were trained to predict, based on simulated AP features in preperturbed cells, whether each cell would develop arrhythmic dynamics in response to each trigger. We found that APD can accurately predict how cells respond to the simple Current Injection trigger but cannot effectively predict the response to I_Kr Block or I_CaL Increase. ML predictive performance could be improved by incorporating additional AP features and simulations of additional experimental protocols. Importantly, we discovered that the most relevant features and experimental protocols were trigger specific, which shed light on the mechanisms that promoted arrhythmia formation in response to the triggers. Overall, our quantitative approach provides a means to understand and predict differences between individuals in arrhythmia susceptibility.

QT Assessment in Early Drug Development: The Long and the Short of It

The QT interval occupies a pivotal role in drug development as a surface biomarker of ventricular repolarization. The electrophysiologic substrate for QT prolongation coupled with reports of non-cardiac drugs producing lethal arrhythmias captured worldwide attention from government regulators eventuating in a series of guidance documents that require virtually all new chemical compounds to undergo rigorous preclinical and clinical testing to profile their QT liability. While prolongation or shortening of the QT interval may herald the appearance of serious cardiac arrhythmias, the positive predictive value of an abnormal QT measurement for these arrhythmias is modest, especially in the absence of confounding clinical features or a congenital predisposition that increases the risk of syncope and sudden death. Consequently, there has been a paradigm shift to assess a compound's cardiac risk of arrhythmias centered on a mechanistic approach to arrhythmogenesis rather than focusing solely on the QT interval. This entails both robust preclinical and clinical assays along with the emergence of concentration QT modeling as a primary analysis tool to determine whether delayed ventricular repolarization is present. The purpose of this review is to provide a comprehensive understanding of the QT interval and highlight its central role in early drug development.

Impact of disease state on arrhythmic event detection by action potential modelling in cardiac safety pharmacology

INTRODUCTION

The use of in silico cardiac action potential simulations is one of the pillars of the CiPA initiative (Comprehensive in vitro Proarrhythmia Assay) currently under evaluation designed to detect more accurately proarrhythmic liabilities of new drug candidate. In order to take into account the variability of clinical situations, we propose to improve this method by studying the impact of various disease states on arrhythmic events induced by 30 torsadogenic or non-torsadogenic compounds.

METHOD

In silico modelling was done on the human myocytes using the Dutta revised O'Hara-Rudy algorithm. Results were analysed using a new metric based on the compound IC_50s against the seven cardiac ionic currents considered to be the most important by the CiPA initiative (I_Kr, I_Ks, I_Na, I_NaL, I_K1, I_to, I_CaL) and the minimal rate of action potential voltage decrease calculated at the early-afterdepolarization (EAD) take-off membrane voltage (V_min).

RESULTS

The specific threshold at which each torsadogenic compounds induced EAD, was exacerbated by the presence of cardiac risk factors ranked as follows: congestive heart failure > hypertrophic cardiomyopathy > cardiac pause > no risk factor. Non-torsadogenic compounds induced no EAD even in the presence of cardiac risk factors.

DISCUSSION

The present study highlighted the impact of pre-existing cardiovascular disease on arrhythmic event detection suggesting that disease state modelling may need to be incorporated in order to fully realize the goal of the CiPA paradigm in a more accurate predictability of proarrhythmic liabilities of new drug candidate.

A web portal for in-silico action potential predictions

Introduction

Multiple cardiac ion channels are prone to block by pharmaceutical compounds, and this can have large implications for cardiac safety. The effect of a compound on individual ion currents can now be measured in automated patch clamp screening assays. In-silico action potential models are proposed as one way of predicting the integrated compound effects on whole-cell electrophysiology, to provide an improved indication of pro-arrhythmic risk.

Methods

We have developed open source software to run cardiac electrophysiology simulations to predict the overall effect of compounds that block I_Kr, I_CaL, I_Na, I_Ks, I_K1 and I_to to varying degrees, using a choice of mathematical electrophysiology models. To enable safety pharmacology teams to run and evaluate these simulations easily, we have also developed an open source web portal interface to this simulator.

Results

The web portal can be found at https://chaste.cs.ox.ac.uk/ActionPotential. Users can enter details of compound affinities for ion channels in the form of IC₅₀ or pIC₅₀ values, run simulations, store the results for later retrieval, view summary graphs of the results, and export data to a spreadsheet format.

Discussion

This web portal provides a simple interface to reference versions of mathematical models, and well-tested state-of-the-art equation solvers. It provides safety teams easy access to the emerging technology of cardiac electrophysiology simulations for use in the drug-discovery process.

Prediction of Thorough QT study results using action potential simulations based on ion channel screens

INTRODUCTION

Detection of drug-induced pro-arrhythmic risk is a primary concern for pharmaceutical companies and regulators. Increased risk is linked to prolongation of the QT interval on the body surface ECG. Recent studies have shown that multiple ion channel interactions can be required to predict changes in ventricular repolarisation and therefore QT intervals. In this study we attempt to predict the result of the human clinical Thorough QT (TQT) study, using multiple ion channel screening which is available early in drug development.

METHODS

Ion current reduction was measured, in the presence of marketed drugs which have had a TQT study, for channels encoded by hERG, CaV1.2, NaV1.5, KCNQ1/MinK, and Kv4.3/KChIP2.2. The screen was performed on two platforms - IonWorks Quattro (all 5 channels, 34 compounds), and IonWorks Barracuda (hERG & CaV1.2, 26 compounds). Concentration-effect curves were fitted to the resulting data, and used to calculate a percentage reduction in each current at a given concentration. Action potential simulations were then performed using the ten Tusscher and Panfilov (2006), Grandi et al. (2010) and O'Hara et al. (2011) human ventricular action potential models, pacing at 1Hz and running to steady state, for a range of concentrations.

RESULTS

We compared simulated action potential duration predictions with the QT prolongation observed in the TQT studies. At the estimated concentrations, simulations tended to underestimate any observed QT prolongation. When considering a wider range of concentrations, and conventional patch clamp rather than screening data for hERG, prolongation of ≥5ms was predicted with up to 79% sensitivity and 100% specificity.

DISCUSSION

This study provides a proof-of-principle for the prediction of human TQT study results using data available early in drug development. We highlight a number of areas that need refinement to improve the method's predictive power, but the results suggest that such approaches will provide a useful tool in cardiac safety assessment.