Already a DIA Member? Sign in. Not a member? Join.

Sign in

Forgot User ID? or Forgot Password?

Not a Member?

Create Account and Join

Menu Back to Poster-Presentations-Details

W-01: Comparison of Predictive Power of ECG Biomarkers for Detection of Drug-Induced Cardiac Ion Channel Block





Poster Presenter

      Brian Brockway

      • President
      • Vivaquant, Inc.
        United States

Objectives

Evaluate power to predict arrhythmia risk for three ECG repolarization metrics (QTcF, JTpc, and TpTe) obtained using four different cardiac interval measurement protocols.

Method

Repolarization intervals were measured on ECG data from FDA Study 1 using two software applications. Predictive power was evaluated using logistic regression and quantified with receiver operating characteristic area under the curve (AUC). Risk thresholds were estimated with decision trees.

Results

The four protocols applied to FDA Study 1 data (Johannesen 2014) to measure intervals were: a) intervals measured on signal averaged ECG complexes derived from each of three 10-sec ECG strips by FDA software at 16 predetermined timepoints (TSAT), b) the average of intervals derived from every normal beat (beat-to-beat - BTB) in each of three 10-sec ECG strips at each of 16 predetermined timepoints (BTBT), c) 5-min BTB averages measured every 30-min over 24 hours (BTB5), and d) 30-min BTB averages measured every 30-min over 24 hours (BTB30). BTB fully automated (no human review or editing) measurements were obtained using Rhythm Express (VivaQuant, St. Paul, MN). For all methods, the data was limited to active drug windows (40% Cmax) over 24 hours. The predictive power of the models computed for TSAT, BTBT, BTB5, and BTB30 were 94%, 96%, 98%, and 99%, respectively. Results show that when intervals were derived from three 10-sec ECG strips (triplicates) by averaging BTB measurements (BTBT protocol), the predictive power was improved by 2% relative to intervals derived from signal averaged beats (TSAT protocol). Further, the use of 5-min and 30-min averages of BTB values was found to improve predictive power by 4% and 5%, respectively, relative to the TSAT protocol. Mean heart rates at baseline were between 68 and 70 bpm for BTB5 and BTB30 protocols and 56-58 bpm for TSAT and BTBT protocols for each study day. For the predictive models evaluated on this dataset, QTcF was found to be superior to JTpc and TpTe in predicting cardiac risk for all four interval measurement protocols. The decision tree models identified the range of uncertainty around QTcF threshold between 17 and 28 msec for TSAT, 21 and 27 msec for BTBT, 8 and 16 msec for BTB5, and 14 and 20 msec for BTB30. The uncertainty regions have higher range and start at higher values for triplicate measurements, reflecting the selection bias of beats with stable and lower heart rate.

Conclusion

This study demonstrated that examination of continuous ECG measurements of QTcF, JTpc, TpTe intervals over 24 hours is possible with highly automated software. The 2% improvement in predictivity observed with BTBT vs. TSAT protocol may be related to a 25% reduction in standard deviation (SD) of interval measurements provided by Rhythm Express relative to FDA measurements. In this dataset, 10-sec triplicates bias the sampling toward slower heart rates and select the beats with the least variation between cardiac cycles (beats that may not be representative of the inherent non-stationarity of QT/RR relationships). Including all beats within 5-min timepoints (BTB5 protocol) improved predictivity by 4% relative to the traditional triplicate measures. Expansion of the time window to 30 minutes (BTB30) further improved predictivity to achieve AUC=.99. These results call into question whether the traditional approach of measuring intervals from signal averaged beats for 10-sec ECG strips during predefined supine rest periods is optimal for cardiac safety assessment. The beats used to form the composite waveform from which interval measurements are derived may not be sufficiently representative of cardiac dynamics to accurately predict the risk of arrhythmia in other autonomic states or physiological conditions. These results are limited by inclusion of the data only from FDA study 1.

Be informed and stay engaged.

Don't miss an opportunity - join our mailing list to stay up to date on DIA insights and events.