Linking topiramate exposure to changes in electrophysiological activity and behavioral deficits through quantitative pharmacological modeling

Abstract

Topiramate is a broad-spectrum anti-epileptic drug used to treat a variety of conditions, including epilepsy, migraine, substance abuse, mood, and eating disorders. We investigated the effects of topiramate on the working memory system using population pharmacokinetic-pharmacodynamic modeling and unsupervised machine learning approaches. Working memory is the capacity-limited neurocognitive system responsible for simultaneous maintenance and manipulation of information in order to achieve a goal. Behavioral and electrophysiological indices of working memory function were measured using data collected during a double-blind, placebo-controlled crossover study in healthy volunteers. Subjects completed a Sternberg working memory task, during which accuracy and reaction time were measured, while subjects’ EEG was recorded. A pharmacokinetic-pharmacodynamic model was constructed which demonstrated that accuracy decreased linearly as a function of plasma concentration, and that the magnitude of individual deficits was predicted by working memory capacity. A separate pharmacokinetic-pharmacodynamic model was developed which showed that spectral power in the theta frequency band (4-8 Hz) recorded during the retention phase of the Sternberg task increased as a function of plasma concentration. Furthermore, a mixture model identified two subpopulations with differential sensitivity in topiramate-induced theta reactivity. In the subpopulation defined by lower reactivity, reaction times were 20% slower than in the high theta reactivity subpopulation. Principal component regression was used to quantify the relationship between changes in multiple measures of electrophysiological activity and behavioral deficits. Theta power during retention was found to be the best predictor of topiramate-related behavioral deficits. Performance on another working memory task, Digit Span Forward, was also predicted by theta power during retention, as well as alpha (8-12 Hz) power during encoding and retrieval stages. In conclusion, two treatment-independent factors that predict differences in behavioral and electrophysiological responses to topiramate administration were identified: working memory capacity and theta reactivity. Future research will be needed to determine the utility of these demographic factors in predicting risk of cognitive side effects in patients eligible for treatment with topiramate.