A method for assessing and monitoring consistency of nonclinical ECG analysis

Abstract

Cardiovascular safety is a key area of concern for new drugs in development, and the collection and analysis of electrocardiograms (ECGs) is a standard and major component of nonclinical testing. Digital data capture technology allows for high-throughput and long-duration ECG collections, resulting in large volumes of data. Consistent analysis of these ECG data is critical for detecting meaningful changes during nonclinical drug development. We developed a method to assess the consistency of nonclinical ECG analysis for a group of analysts over time. Eight‑lead ECGs were collected from conscious dogs using Ponemah (v5.2, DSI). Analysts placed Pstart, Qstart, Rpeak, Send, and Tend marks on six waveforms for each animal. The ECG files were randomized and re-marked under blinded conditions 4 to 14 days following initial mark placement. Averages of each parameter measured (RR interval, QRS duration, PR interval, and QT interval) were compiled for each marking session and analyst. A Gage R&R evaluation was completed. Graphical output from the Gage R&R evaluation showed distinct variability on group and individual analyst levels. Differences in inter- and intra-analyst variability (reproducibility and repeatability, respectively) were observed between trained analysts and analysts in training. The Gage R&R method is an effective tool for assessing consistency of digital ECG mark placement at a group level. Furthermore, it is able to identify areas of improvement for individual ECG analysts and to assess ECG analyst consistency during their training period. The assessment results are useful for facilitating discussions on best practices and maintaining consistency of mark placement.