Session P36.5

Two Probabilistic Methods to Characterize and Link Drug Related ECG Changes to Diagnoses from the PTB Database: Results with Moxifloxacin

R Bousseljot, D Kreiseler, S Mensing, A Safer*

Drug Safety Consultant
Weisenheim am Sand, Germany

Safety assessment of drugs for risk of severe ventricular arrhythmias is currently performed by analysis of the QT interval. Unfortunately prolongation of QT is a weak predictor, burdened by the disadvantage of not having direct link to any clinical endpoint.
We propose two methods that might bridge the gap, using about 30, 000 ECGs from the PTB database. Each ECG is accompanied by well defined diagnoses by two or more cardiologists. Both methods require ECG signal preprocessing: selection of one significant „representative“ beat from each 10sec ECG. After PQRST recognition, heart rate scaling (Fredericia correction) is applied.
Method A is based on the assumption that two ECGs from the same person show limited variability if no influential external factors impact. Probabilistic intra-subject clustering of 12-lead based waveform similarities is applied. Clusters under baseline condition are compared with placebo and treatment clusters as well as baseline and treatment clusters between study periods.
Method B is based on the assumption that two ECGs, whose multi-lead waveforms are in best match, have the same diagnoses. After preprocessing, all 12-lead waveforms are compared with the full PTB database heart rate normalized raw waveforms, and assigned a probability for normality and a limited number of cardiac disease diagnoses. The arrays of disease probabilities are compared between treatments and periods.
Material of investigation for proof-of-concept was a single-blinded placebo controlled crossover design phase-1 study with 10 volunteers measured over 2 periods with 3 days each over 7 time points/day with triplicate resting ECGs/time point. Volunteers were treated with iv isotonic saline, 200 and 400 mg Moxifloxacin. This drug is considered as a positive control in thorough QT studies. It prolongs the QT interval, but shows low risk to act proarrhythmic. Therefore the expected outcome is a significant QT prolongation without an increase in proarrhythmia probability.
Study results for QT prolongation showed an average „double-delta“ QT difference of 6.7 msec for 200 mg and 13.5 msec for 400 mg Moxifloxacin treatment.
Results for method A showed larger similarities of ECG patterns (based upon Mahalanobis distance) between baseline and placebo conditions than under treatment conditions. In just one case out of ten no difference between placebo and treatments could be found
Method B results showed no significant treatment related changes in terms of diagnostic probabilities. This confirms the expected outcome as the arrhythmic propensity is low for Moxifloxacin. A positive verification of the concept by a proarrhythmic drug like Sotalol will be needed to confirm the diagnostic potential of the proposed methods.

(Abstract Control Number: 67)