Abstract
Drug-induced convulsions have a significant impact on drug development, but no reasonable biomarker has been identified yet. The previous study showed the index generated by heart rate variability (HRV) in monkeys using machine learning could be a biomarker for GABAA receptor antagonist-induced convulsion. In the present study, we tried to apply this method to other convulsants. We also examined non-convulsants affecting the autonomic nervous system to verify any false positives. Telemetry-implanted male monkeys were dosed with convulsants (4-AP, kainic acid, ranolazine, and bupropion) at some dose levels, and their electrocardiograms (ECGs) were recorded. The ECG data during the predose period were used as the training data, and HRV was analyzed by multivariate statistical process control. As a result, the index (Q statistic) of 4-AP increased at lower than the convulsive dose. Kainic acid and ranolazine also increased the index at the convulsive dose. Bupropion did not change the index up to the highest dose, 1/3 of the convulsive dose. The same analysis was applied to non-convulsants (atropine, atenolol, and clonidine) and clear changes in the index were observed. The elevation of the index accompanied or preceded changes in autonomic nerve activity (HF and LF/HF), suggesting the index did not reflect convulsion potential directly but detected changes in the autonomic nerve activity sensitively. Although there is a possibility of false positives, this method is useful to predict drug-induced convulsions when considering the pharmacological profile of the compound.
