Authors :
Presenting Author: David Geng, BS – University of Pittsburgh School of Medicine, Carnegie Mellon University
Shobhit Singla, MD, PhD – University of Pittsburgh School of Medicine Dept. of Neurology
James Castellano, MD, PhD – University of Pittsburgh School of Medicine Dept. of Neurology
Avniel Ghuman, PhD – University of Pittsburgh School of Medicine Dept. of Neurosurgery, Carnegie Mellon University
Rationale:
Current knowledge of anti-seizure medications (ASMs) largely derives from comparing those on ASMs to those not receiving them, leaving gaps in understanding the continuous, individualized relationship between neural activity and ASM pharmacodynamics and the mechanisms of ASM effects. This study assessed the effectiveness of using scalp electroencephalography (EEG) to track pharmacodynamics and identify key EEG markers predictive of ASM blood levels.Methods:
We analyzed retrospective continuous EEG recordings from 14 patients, each monitored for 3 to 9 days in the EMU for phase 1 evaluation for epilepsy and later diagnosed as non-epileptic. During their stay, each patient received at least one oral dose of a single ASM (levetiracetam, topiramate, lamotrigine, oxcarbazepine, or divalproex sodium).
EEG Feature Extraction: EEG data were segmented into 5-second non-overlapping epochs after thorough preprocessing and artifact removal. We extracted 21 time and frequency domain features from each of the 28 electrodes, totaling 588 features per epoch.
Pharmacokinetic Modeling: We developed a two-compartment model for each patient to simulate drug blood levels over time, incorporating drug concentration at admission, dosage schedules, and established pharmacokinetic parameters.
Regression Analysis: Ridge Regression, out of sample cross-validation of predictions of pharmacokinetic levels based on brain activity, and permutation testing were used to analyze the relationship between EEG features and drug levels, ensuring that the results were robust and predictive outside of the regression training sample.
Results:
Our analysis demonstrated a statistically significant relationship between EEG-derived predictions and ASM drug levels in 12 out of 14 subjects at the p = 0.05 level, with 8 subjects reaching significance at the p = 0.01 level. This finding underscores the potential of EEG-based models to reliably predict pharmacodynamic responses. Notably, aperiodic EEG features, linked to background activity and the excitatory-inhibitory (E/I) balance, consistently emerged as the top predictors across all subjects, regardless of statistical significance. This consistency suggests that these features are robust markers of pharmacodynamic effects and crucial for tracking ASM blood levels across diverse patients.
Conclusions:
Our findings show that scalp EEG signals can reliably serve as pharmacokinetic predictors, offering a non-invasive method to monitor neural dynamics in response to ASMs. The consistent importance of aperiodic features supports the hypothesis that ASMs modulate the E/I balance and EEG background activity. We will expand our dataset to include patients with epilepsy to determine whether these findings generalize across both non-epileptic and epileptic populations to assess how these E/I changes relate to seizure suppression. This research is significant in that it paves the way for using scalp EEG in continuous and long-term pharmacokinetic applications, potentially improving the understanding of ASM effects on large-scale brain networks.Funding:
R01MH132225