While stress-induced seizures can spread as cortical paroxysms, not all convulsions triggered by stress are associated with cortical activation. A subset of these events may instead fall under the category of psychogenic nonepileptic seizures (PNES).

Clinically, PNES (or Functional Neurological Disorder - FND) are a condition at the intersection of neurology and psychiatry, presenting as convulsive events associated with heightened emotional reactivity, and thought to affect the regulation of motor output (Diez et al., 2021), (Boutros et al., 2019). PNES are considered the manifestation of an acquired network disorder related to emotional learning/reactivity and motor control circuitry involving limbic regions (Amiri et al., 2021; Balachandran et al., 2021). PNES are currently considered a multi-factorial condition involving complex interactions between environmental, genetic, and psychological factors, with experiences of childhood trauma being the most common risk factor (Yang et al., 2023; Goldstein et al., 2019).

PNES lacks ictal epileptiform discharges or seizures on EEG, although the EEG may show other abnormal nonspecific changes; they are resistant to traditional antiseizure drugs; and at least partially responsive to induction or provocative techniques. Although PNES convulsions do not display abnormal cortical synchrony, their repeated presence can trigger secondary paroxysms (sudden, transient bursts of abnormal EEG activity, e.g., seizure events), highlighting the intricate interplay between subcortical stress pathways and seizure manifestation (Benbadis et al., 2001; Dickinson and Looper, 2012). In epilepsy clinics, PNES are detected at a relatively high percentage varying from 10% to over 50% (Benbadis, 2006; Benbadis et al., 2001).

PNES are confirmed by simultaneous video and EEG in a neurology setting, confirming an absence of ictal EEG changes and the presence of normal awake EEG rhythms before, during, and immediately after the event (Benbadis, 2006; LaFrance et al., 2013). Activation procedures are critical for the differential diagnosis of PNES (Benbadis, 2006), and are a common intervention for psychogenic movement disorders (Brown, 2006). No single biomarker successfully differentiates PNES from ES, although current efforts are underway to identify additional clinical chemistry markers differentiating between PNES and epileptiform variants (Sundararajan et al., 2016).

Reports from Contract Research Organizations (CROs) conducting routine toxicology studies (Gauvin et al., 2018; Satomoto et al., 2012) where seizures are recorded based on behavioral observation alone, without EEG or detailed Racine scale staging (Racine, 1972) confirm the presence of conditioned stress-induced convulsions with PNES features in rodents, elicited ususally by reactivity to handling, fearful situations, unexpected light changes, or sound. These conditions are commonly associated with dosing or room observations.

Stress convulsions in rats share features with PNES in humans, i.e., acquired sensitivity to environmental cues paralleling trauma, and a strong association with triggers and autonomic dysregulation. The presence of overlapping non-convulsive behaviors such as stereotypy, conditioned responses to cues, and the absence of postictal suppression usually allows for a functional comparison. Given these shared features, PNES terminology can be applied to rodents to describe stress-induced conditions representing maladaptive motor phenomena (Gauvin et al., 2018).

Rat stereotypy and PNES variants are on the same continuum of behaviors sharing mechanistic pathways. Stereotypy in rats is a well-documented non-seizurogenic phenomenon, with invariant behaviors such as excessive grooming, jumping, circling, route-tracing, biting, pacing, shaking, twitching, paddling, etc., often observed in captive environments (Garner, 2005). Rodent stereotypy measurement has a long history in the field of psychopharmacology (Kelley, 2001), first used to describe the physiological effect of stimulant or excitatory drugs like apomorphine or amphetamine, yet spontaneous stereotypy was also documented as a result of stress (Powell et al., 1999).

Mechanistically, stereotypy is believed to involve dopaminergic dysregulation within the basal ganglia and the nigrostriatal or mesolimbic pathways responsible for motor control and reinforcement learning (Powell et al., 1999). Additionally, glutamatergic and GABAergic imbalances contribute to the persistence of these behaviors by impairing inhibitory control over motor circuits. In rodents, stereotypy is often linked to dopaminergic dysregulation within the striatum, where excessive activation leads to rigid, compulsive motor patterns (Canales and Graybiel, 2000), and exacerbated by chronic stress exposure (Garner, 2005).

Stereotyped non-epileptiform hyperkinesis presents rhythmic side-to-side swings of the head and torso and pronounced pendulum-like movements, abortive or repetitive movements, or stereotyped jumps or shakes (Kelley, 2001; Powell et al., 1999; Garner, 2005). The alternating nature of the movement during stereotypies often appears sufficiently similar to clonic–tonic convulsions where rhythmicity of movements is elicited in alternations between muscle contractions and relaxations. To complicate matters further, stereotypy can occasionally be present as an intermediate phase when a shift from a catatonic to an epileptiform seizure occurs (Ryazanova et al., 2023), and the association of these events in sequence may add to the difficulty of distinguishing between them.

In general, stereotypies are pathological motor responses to environmental factors involving dysregulation of basal ganglia, limbic system, or the prefrontal cortex, which regulate movement, emotional responses, and behavioral flexibility. These phenomena can be viewed through the lens of learned or conditioned responses, developing when an animal repeatedly engages in a behavior that provides a temporary coping mechanism and reinforced even when it is no longer adaptive.

While stereotypies/PNES and drug-induced seizures arise from distinct neural mechanisms, they exhibit sufficiently similar behavioral features and clinical observations. However, since stereotypy can manifest with or without underlying epileptiform activity, it is not a reliable prodromal indicator of seizure when stress conditioning is suspected.

Preclinical studies introduce multiple environmental stressors that can function as triggers/cues for conditioned responses and can vary between facilities.

Dosing is a stress factor and can induce anticipatory stress responses, particularly when paired with aversive handling techniques, or with restraint (Balcombe et al., 2004). Routine laboratory noise, across humanly audible range or ultrasonic, can also act as an unexpected stressor (Turner et al., 2005). Because rodents are nocturnal, facility activities occurring during the day are inherently more stressful. Many of these factors resemble cued fear conditioning paradigms (Buccafusco, 2009), variants of the stress response models which can alter the natural freezing response to both contextual and auditory conditional stimuli and participate indirectly to disinhibition of repetitive movements (Goosens and Maren, 2001).

Variations in social housing, bedding material, and cage type can also impact stress resilience (Van Loo et al., 2003). Rodents are highly social animals, and isolation has been shown to increase anxiety behaviors and dysregulate stress hormone levels. Social buffering (Kiyokawa et al., 2004), where the presence of cage mates helps dampen stress responses, is a well-documented phenomenon that can reduce the likelihood of stress conditioning.

Overall, the elements of a conditioning context require features bound into a conjunctive representation, sharing mechanisms common to formation of memories (Rudy et al., 2004). Dissociating learning due to the cue and the context adversity is difficult, as in some cases the stress related to the cue itself may also interfere with contextual learning (Buccafusco, 2009).

Inter-facility differences in husbandry, environmental enrichment, and personnel training can significantly affect stress exposure (Gauvin et al., 2018). Therefore, when variability in the incidence of convulsions across facilities is observed, it may indicate underlying stress conditioning.

In regulatory general toxicology studies, behavioral seizures are typically described in general terms such as “clonic,” “tonic-clonic,” or “convulsion,” often without further phenotypic classification. No mandatory standardized behavioral scoring system is routinely applied, (although general parameters of duration may be included), and observational criteria may vary across facilities depending on local procedures. The Racine scale widely cited in epilepsy research and developed for induced seizure models in preconditioned animals is not applied to neurologically normal animals in toxicology settings. This lack of resolution of general toxicology study convulsion observations contributes to the ambiguity of convulsive findings.

Since stress-induced convulsions and drug-induced seizures may present with similar motor manifestations, their differentiation requires a systematic approach appropriate for toxicology settings as follows:

If stress-conditioning is suspected in rat while all other preclinical and clinical data (including EEG, alternate species, or human safety data to date) remain seizure-free (or occurring at high multiples exposure in the severe toxicity range), the rat convulsive events have a high likelihood of being PNES and relying on more predictive species may be justified.

This distinction is particularly important when canine toxicology data show a promising CNS safety profile, as dogs are inherently seizure-sensitive (Berendt et al., 2015; Forsgård et al., 2019; DaSilva et al., 2020). However, the implications of using dogs in seizure liability assessments differ from those of stress-induced PNES in rats. Unlike rats, dogs are susceptible to true idiopathic epilepsy that may be triggered at lower doses of an investigational drug with actual proconvulsant properties.

We present these simple distinguishing criteria specifically because routine toxicology studies are not designed to differentiate seizure subtypes. Animals are normal and not seizure-prone models, electroencephalographic data are rarely collected, and convulsive events are typically documented based on gross behavioral observation. As a result, behaviors consistent with stress-induced stereotypy or PNES may be misclassified as epileptiform seizures. The framework we propose is designed to bring greater structure and interpretive clarity to these ambiguous observations. Importantly, it offers feasible, low-burden strategies such as reflex testing, context evaluation, and timing relative to procedural stressors, that can be readily implemented in ongoing studies to improve confidence in seizure liability assessments.

From a regulatory and scientific standpoint, EEG remains the definitive tool for confirming electrographic seizure activity, and differentiating PNES from ES. However, in routine toxicology studies, EEG is rarely implemented, and animals are not seizure-prone models. Consequently, events are recorded behaviorally without electrographic validation, and standardized seizure staging systems such as the Racine scale developed for controlled, induced seizure paradigms is not fesible in this context. It is only in the context of highly resource-intensive follow-up studies in rodent that EEG is usually deployed.

Moreover, when stress conditioning is suspected, it is also important to recognize both the value and limitations of EEG measurements in additional focused studies. While EEG is the gold standard for differentiating between ES and PNES, both can be triggered by stress in the absence of a drug effect. EEG does not inherently differentiate whether the observed response is due to a direct drug effect or a conditioned stress response and should be used selectively, when distinguishing between these two mechanisms provide meaningful insight into the interpretation of seizure risk (e.g., when PNES alone is suspected).

It is essential to distinguish these scenarios from the use of EEG as a standard method for assessing drug-induced seizure risk. When no substantial evidence suggests stress conditioning in rats, EEG remains a crucial tool for identifying drug-induced seizure risk and determining a No Observed Adverse Effect Level (NOAEL).

Given the impact of convulsions in non-clinical studies and the confounding potential of stress-related conditioning in rodents, there is a need for greater awareness and operational preparedness among facilities conducting rat studies. Significant improvements could be achieved by ensuring test facilities are prepared to detect stress conditioning and differentiate between ES and PNES through distraction techniques, assessing associations with potential triggering cues (e.g., light-dark cycle interruptions, handling prior to dosing), and evaluating their relationship to the maximum observed concentration of a drug in plasma following administration, or C_max. Late-onset convulsions are not uncommon in chronic rat toxicity studies and may arise from multiple factors, including the test article, age, stress conditioning, or a combination thereof. Even when a definitive conclusion cannot be drawn from a single study, it is recommended that study directors routinely document and contextualize potential stress conditioning effects, providing a more comprehensive and transparent foundation for regulatory discussions.

It is of value to discern stress-related convulsions, from those related to severe systemic toxicity. Kidney toxicity can lead to uremia, metabolic acidosis, or electrolyte imbalances (e.g., hyponatremia, hyperkalemia, hypocalcemia). Gastrointestinal (GI) toxicity can cause dehydration, nutritional deficits, or metabolic imbalances and systemic inflammation, all of which lower the seizure threshold by altering neuronal excitability. These factors should be evaluated in a WoE framework to ensure a more accurate attribution of convulsive risk.

For programs under the US FDA/CDER, restrictions on human exposure may be imposed on first-in-human trials (FDA, 2005) and further limitations on maximum clinical doses as outlined in ICH M3 (R2) (ICH M3, 2009) and its Q&A (ICH guideline M3, 2012). This typically leads to additional safety margin restrictions on both AUC and C_max, often limiting human exposure to 1/10th of the NOAEL for convulsions in the most sensitive species, even in the absence of such events in humans. Other regulatory agencies, such as the MHRA and EMA, may place greater emphasis on clinical experience when human data are available, requesting clinical dose escalation modifications or enhanced monitoring.

When stress-induced events are suspected in rats, we propose adopting a Weight-Of-Evidence (WoE) framework to assess the presence of PNES, considering events in all species, the test agent’s on- and off-target effects relevant to seizure mechanisms, the susceptibility of the clinical trial population, and the fesibility of detecting prodromal activity in humans. Alternative species should be emphasized when multiple indicators of rodent stress conditioning are present. Specifically, if convulsions exhibit PNES features, facility-dependent variability, lack postictal events or pathology findings, and are either absent or occur only at substantially higher exposure multiples in other species, they should be viewed as suspect.

Failing to remove bias introduced by stress into safety assessments can lead to false-positive seizure liability findings that may result in premature termination of otherwise viable drug candidates. Parsing PNES from epileptiform events is essential to ensure that decision-making reflects true translational risk. When the root cause of convulsive events is stress-conditioning rather than pharmacologic neurotoxicity, recognizing this distinction prevents the misallocation of resources, prolonged delays, and unnecessary animal use, while preserving confidence in drug safety. While additional non-clinical investigations may sometimes be necessary, in many cases a targeted proportionate approach can safeguard clinical subjects and address regulatory concerns without excessive delays, resource expenditures, and increased animal use.

We believe that by applying a WoE to convulsion observations, a more appropriate non-clinical and clinical investigative program can be developed without sacrificing safety for clinical subjects and, leading to a more rapid and efficient drug development of clinically important therapeutics.

Table 1 outlines a first attempt towards key considerations that can increase or decrease the relevance of rodent convulsions to human risk assessment.