Predicting drug-mediated pro-arrhythmic effects using pre-drug electrocardiograms. (July 2021)

• Record Type: Journal Article
• Authors: Peng, Tommy; Malik, Avinash; Trew, Mark L.
• Is Part Of: Biomedical signal processing and control. Volume 68(2021)
• Publication Date: 2021-07
• Subjects: Electrocardiogram decomposition -- Pro-arrhythmic risk -- Gaussian Mesa Functions -- Electrocardiogram prediction; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.102712
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Electrocardiograms (ECGs) are decomposed using Gaussian Mesa Functions (GMFs). Population trends in drug induced ECG changes are captured using GMF parameters. Post-dose GMF parameter values are predicted using trends and pre-dose GMF values. Predicted post-dose JT, QT, T peak to T end intervals are similar to expert annotations. Predicted post-dose GMF parameters produce realistic post-dose ECG morphologies. Abstract: Altered electrocardiogram (ECG) morphology is important for assessing cardiac pro-arrhythmic risk of drugs. We propose a basis function method to predict morphological and QT, JT, Tpeak—Tend timing interval changes in ECGs due to drug effects. The method systematically decomposes ECGs for a study population recorded at varying pharmacokinetic states into Gaussian Mesa Functions (GMFs). The GMF parameters are then fit to linear mixed effects drug sensitivity models. For a new subject, post-drug GMF parameter values and ECG morphology at varying pharmacokinetic states are predicted from the pre-drug GMF parameter values using the drug sensitivity models. The proposed methodology is validated with clinical ECGs of human subjects administered Dofetilide, Quinidine, Ranolazine, and Verapamil. The datasets are obtained from the ECGRVDQ database. The proposed method predicts post-drug timing intervals not significantly different to expert annotated intervals (pair-wise t -test p > 0.05 ) for 153 out of 180 scenarios (drug type, hours post-dose, and ECG timing interval combinations). A comparative method based on expert annotations predicts 105 out of 180 scenarios. Importantly, realistic predictions of post-dose ECG morphology are reconstructed from predicted GMF parameters. Our study suggests that the GMF parameter space can provide important new biomarkers for assessing and visualizing drug-induced changes in ECGs.