Rethinking preclinical models in drug-resistant epilepsy: From seizures to side effects all at once

Abstract

Commentary on “Challenging the preclinical paradigm: Adverse effects of antiseizure medicines in male rats with drug-resistant epilepsy” by Guignet et al., British Journal of Pharmacology (2025); doi: 10.1111/bph.70076.

Drug-resistant epilepsy (DRE) remains a major unmet need in the area of epilepsy research (Chen et al., 2018); it affects nearly one-third of people with epilepsy despite the availability of over 30 antiseizure medications (ASMs), with most being authorized in the last 25 years (Chen et al., 2018). The persistence of this clinical burden raises critical questions about the translational validity of preclinical models traditionally employed in the discovery and evaluation of ASMs. This study by Guignet et al. (2025), and others from the same group (Barker-Haliski & White, 2020), further underline this problem while proposing and executing a paradigm change in how ASMs should be evaluated in animal models. Specifically, here, the first shift in the preclinical approach is based on moving from acute seizure induction in neurologically intact animals to chronic drug exposure in disease-relevant models that more accurately mimic human pharmaco-resistance (Guignet et al., 2025).

Furthermore, this study addresses a crucial gap in epilepsy research: the inadequacy of existing preclinical methodologies to predict efficacy and tolerability of ASMs in drug-resistant forms of epilepsy at the same time. In conventional screening models such as the maximal electroshock seizure (MES) and pentylenetetrazol (PTZ) tests, seizures are acutely evoked in otherwise healthy rodents, and compounds are evaluated for their ability to suppress these events. While these models are efficient in screening for antiseizure potential, they fail to capture the chronic and multifaceted nature of epilepsy, particularly the pathophysiological, pharmacokinetic and pharmacodynamic complexities inherent in DRE.

Guignet and colleagues present a refined approach using a kainic acid-induced status epilepticus model leading to spontaneous seizures, a well-established and widely used model of temporal lobe epilepsy in rodents. They administer ASMs via a chronic, steady-state oral dosing regimen embedded within food pellets, mimicking patients' real-world treatment. They further integrate continuous video-EEG monitoring, pharmacokinetic profiling and behavioural tolerability assessments, which are often neglected in preclinical epilepsy research but relevant for a meaningful clinical translation.

Three widely used ASMs (carbamazepine [CBZ], levetiracetam [LEV], and lamotrigine [LTG]) were tested in a cross-over design, allowing each animal to serve as its own control. This design is elegant and statistically robust, reducing inter-animal variability and allowing direct comparisons across treatments and doses. By including therapeutic drug monitoring, the researchers ensure that observed effects (or lack thereof) are not simply attributable to inadequate exposure, but reflect pharmacodynamic outcomes.

Both CBZ and LEV demonstrated a modest but consistent reduction in seizure frequency (≥50%) in about half the animals, with some achieving complete seizure freedom. These results were dose-independent, suggesting a plateau in efficacy and potentially a ceiling effect in this chronic model. Plasma drug levels for both agents were within or slightly above clinically relevant ranges, affirming the translational relevance of these findings.

High-dose CBZ was associated with motor impairment, a side effect observed clinically as ataxia or sedation, especially at higher doses or in polytherapy. LEV, on the other hand, was well tolerated and did not affect weight, food consumption or motor behaviour.

On the other hand, at doses exceeding the therapeutic range, LTG not only failed to reduce seizures but significantly worsened seizure frequency and in cluster size and duration. Furthermore, LTG also induced profound behavioural changes: hyperexcitability, aggressive responses to handling and reduced food intake leading to weight loss. While the mechanisms underlying these effects remain speculative, the authors rightly point to the need for further investigation into the pharmacodynamic alterations in drug-resistant brain tissue, where target expression or function may be altered.

The authors' use of a chronic epilepsy model with spontaneous seizures, seizure clusters and pharmacoresistant features represents a critical advancement. Their findings underscore that ASM efficacy and tolerability cannot be divorced from the biological context in which they are tested. In fact, many of the limitations in ASM development may stem from testing drugs in models that lack the main features (e.g., neuroinflammation, altered receptor expression and network reorganization) that drive treatment resistance in patients.

Moreover, the study draws attention to the pharmacokinetic–pharmacodynamic disconnect often encountered in ASM therapy, further supporting the need for therapeutic drug monitoring (TDM) in optimizing ASM use.

The authors conclude that, while the paper focuses on three older-generation ASMs, the broader implications of this work are substantial. This platform could be employed to test novel compounds, combination therapies or even interventions aimed at disease modification. It could also serve as a foundation for precision medicine strategies, by identifying responders and non-responders within the same population, potentially guided by biomarkers.

Despite its strengths, the study does have limitations. It includes only male rats, leaving open the question of sex differences in pharmacoresistance and ASM tolerability, a highly relevant issue given known sex-specific pharmacokinetics and neurobiology. Future studies should extend this paradigm to female animals and ideally to multiple epilepsy models (e.g., genetic epilepsies, post-traumatic epilepsy) to assess generalizability. Another limitation is the lack of measurement of brain ASM concentrations, which could provide crucial information about drug penetration and target engagement.

The cross-over protocol with two doses of each drug is very elegant but must keep under consideration eventual carry over effects or slow pharmacokinetics. For experimental drugs, such parameters are not always completely known and drug–drug interactions are extremely frequent in epilepsy management (Roberti et al., 2025).

The chosen model is time-consuming and requires an infrastructure, therefore is not completely manageable or rapid for potential ASM screening; acute models were used to this purpose in the past and still now.

Finally, cognitive and affective side effects, which are critical determinants of ASM success in patients, were not assessed. Incorporating behavioural and cognitive batteries of tests into future studies would enhance the translational value of this approach.

Guignet et al. make a compelling case for updating the preclinical pipeline in epilepsy drug development. Their study demonstrates that the context of drug administration based on chronic dosing, pharmacoresistant tissue and behavioural tolerability, has important influences on drug efficacy and safety. It shows that the same drug can be beneficial, inert or harmful depending on disease stage, drug levels and likely model used. This is based on several important aspects including the pathophysiology underlying seizure generation and a known clinical factor determining the best ASM choice. Genetics and the related animal models today represent an important tool for drug development (Bertocchi et al., 2024).

As the field of epilepsy therapeutics continues to evolve, particularly with the emergence of targeted therapies and gene-based interventions, there is an urgent need for preclinical models that more accurately reflect the clinical reality of DRE. This study provides a robust framework to fill that gap, offering a more faithful translation of ASM performance from bench to bedside.

In the era of precision medicine, where the goal is to tailor treatments to individual patients, preclinical models must be improved to meet this challenge. The work by Guignet et al. represents a crucial step in that direction.

None.