While treatment-emergent sleep disturbances associated with ASMs have been occasionally reported, a comprehensive, large-scale characterization of these events across different ASMs has been lacking. Our study addressed this gap by conducting a systematic, real-world pharmacovigilance investigation of the FAERS database. We provided a class- and drug-specific profile of sleep disturbances in PWE, identifying distinct reporting signals and substantial heterogeneity in clinical manifestations across 19 commonly used ASMs. This analysis offers a foundational framework for understanding the spectrum of ASM-associated sleep disturbances in clinical practice.

This study systematically evaluated major categories of ASMs, including sodium channel modulators, GABAergic system modulators, calcium channel modulators, SV2A modulators, and multi-target agents. Notably, the analysis revealed that even drugs with similar or identical mechanisms of action exhibited distinct associations with sleep disturbances, underscoring the importance of drug-specific evaluation. Among sodium channel blockers, carbamazepine, lacosamide, and lamotrigine generally demonstrated no significant signals of disproportionate reporting for sleep disturbances, consistent with previous literature. Specifically, the impact of carbamazepine on sleep structure was primarily characterized by transient AEs during initial treatment, with longitudinal studies confirming that both sleep macrostructure and daytime alertness returned to baseline levels within 1 month of treatment (Nayak et al., 2016; Jain and Glauser, 2014). Furthermore, clinical studies consistently demonstrated excellent sleep safety profiles for lacosamide. In randomized controlled trials involving both healthy subjects and adults with focal epilepsy, lacosamide showed no clinically significant AEs on polysomnography-measured sleep architecture, subjective daytime sleepiness, or sleep quality (Foldvary-Schaefer et al., 2017). Importantly, some parameters even showed stability or modest improvement, confirming its ability to maintain sleep-wake cycle integrity while controlling seizures (Foldvary-Schaefer et al., 2017; Lupo et al., 2023). Additionally, lamotrigine exhibited predominantly positive effects on sleep architecture in PWE (Legros and Bazil, 2003). Polysomnographic studies clearly demonstrated that adjunctive therapy significantly prolonged rapid eye movement (REM) sleep duration while reducing sleep stage transitions and arousal indices, thereby effectively enhancing sleep continuity and stability. Crucially, these improvements were not accompanied by significant increases in daytime sleepiness (Bazil, 2017; Kataria and Vaughn, 2016; Foldvary et al., 2001; Jain and Glauser, 2014). Overall, lamotrigine demonstrates favorable sleep safety characteristics by improving nocturnal sleep architecture while maintaining daytime alertness levels.

In contrast, our analysis indicates that phenytoin use shows no significant association with sleep disturbances. However, available evidence reveals that impact of phenytoin on sleep architecture demonstrates complexity and multidimensional characteristics. Specifically, some studies report reduced N1 sleep and enhanced slow-wave sleep (Staniszewska et al., 2017), while others observe increased N1 sleep along with suppression of both slow-wave and REM sleep (Lanigar and Bandyopadhyay, 2017). Similarly, we also observed divergent results with phenobarbital (Iivanainen and Savolainen, 1983; Wolf et al., 1984; Liguori et al., 2021). These discrepancies may be closely related to differences in study design, population characteristics, and pharmacokinetic parameters, reflecting the complexity of the neurophysiological effects for drugs and interindividual variability. Meanwhile, the analysis results for oxcarbazepine and topiramate warrant attention. Neither drug showed a significant reporting of sleep disturbances in our findings. However, detailed literature analysis reveals that oxcarbazepine exhibits definite sedative properties, characterized by shortened sleep latency, increased total sleep time and REM sleep, along with higher rates of reported daytime somnolence (Ayala-Guerrero et al., 2009; Thelengana et al., 2019). In contrast, although current research on sleep effects of topiramate remains relatively limited, available evidence suggests that for newly diagnosed epilepsy patients, low-dose, slowly titrated topiramate monotherapy can effectively control seizures without significantly affecting daytime alertness, demonstrating good clinical tolerability (Bonanni et al., 2004). Particularly noteworthy was our finding that eslicarbazepine showed a signal of disproportionate reporting of sleep disturbances, which differs from the conclusion reached by Romigi et al. (2020) based on polysomnographic improvements in sleep microstructure. This discrepancy may stem from differences in assessment dimensions (subjective experience versus objective parameters), study population characteristics, and medication background factors. It should be clarified that our findings do not negate previous conclusions but rather suggest the drug may have dual effects on sleep: potentially improving sleep stability while possibly causing subjective sleep quality deterioration in some patients. Therefore, this hypothesis requires further validation through future trials incorporating both subjective and objective measures in large sample studies. Regarding other ASM classes, our analysis clearly confirms particularly strong associations between sleep disturbances and both stiripentol and cenobamate. Specific data from pre-clinical trials of stiripentol show that 67% of the treatment group reported somnolence symptoms compared to only 23% in the placebo group, with insomnia rates of 12% and 7% respectively (US Food and Drug Administration, 2022). Similarly, cenobamate also causes definite somnolence adverse reactions, requiring particular attention and monitoring during clinical use (US Food and Drug Administration, 2025). The impact of valproic acid on sleep architecture remains controversial. Our results show no statistically significant association between valproate therapy and sleep disturbances. Research on valproic acid reveals a complex and contradictory profile of its effects on sleep architecture. In studies on juvenile myoclonic epilepsy (JME), valproic acid monotherapy has been associated with both beneficial changes, such as increased sleep efficiency, prolonged N2 and N3 sleep, and extended REM duration, and neutral outcomes, such as unchanged sleep latency or daytime sleepiness (Nayak et al., 2015). Conversely, a study in mixed epilepsy types reported consistently disruptive effects, including prolonged N1 sleep, shortened REM sleep, and decreased sleep efficiency, particularly with long-term use (>3 months) (Zhang et al., 2014). This paradox is further highlighted in childhood absence epilepsy (CAE), where sleep outcomes were directly tied to seizure control, improving in fully controlled patients but worsening in those with partial control (Glauser et al., 2013). These findings collectively suggest that the impact of valproic acid on sleep is not uniform but is significantly influenced by epilepsy type, treatment duration, and therapeutic response.

Both brivaracetam and levetiracetam target synaptic vesicle protein 2A, whereas perampanel is a selective AMPA receptor antagonist. Notably, our study found a significant association between brivaracetam and sleep disturbances based on disproportionality analysis, which was not observed for levetiracetam and perampanel. This finding is corroborated by evidence from randomized controlled trials of brivaracetam, in which approximately 20% of participants reported daytime somnolence (Klein et al., 2015). In contrast, some clinical studies have suggested potential sleep-improving effects (Klein and Bourikas, 2024; Faught et al., 2024). However, the methodological limitations and considerable sample heterogeneity in these investigations necessitate cautious interpretation of such results. Overall, the convergence between our real-world pharmacovigilance data and previous controlled trial evidence strengthens the validity of the observed safety signal for brivaracetam. Nevertheless, further studies employing standardized methodologies and larger cohorts remain essential to fully elucidate the safety profiles of these drugs on sleep. In contrast, levetiracetam demonstrates a clear dosedependent effect on sleep. Research indicates that a gradually titrated regimen reaching 2000 mg daily can significantly improve sleep quality, manifested as extended total sleep time, improved sleep efficiency, and enhanced deep sleep (Cicolin et al., 2006; Chaneva, 2021; Thelengana et al., 2019). Importantly, these improvements occur without compromising daytime alertness, suggesting good tolerability and potential sleep-enhancing properties. Meanwhile, perampanel shows positive effects in improving sleep for epilepsy patients. Studies confirm that adjunctive perampanel treatment not only objectively optimizes sleep architecture by increasing deep sleep proportion and improving sleep efficiency, but also significantly enhances subjective sleep quality and alleviates insomnia symptoms without causing daytime drowsiness (Rocamora et al., 2020; Bergamo et al., 2025). For gabapentin, or zonisamide, our study observed no significant associations with sleep disturbances, consistent with previous research findings (Romigi et al., 2021; Carvalho et al., 2022). However, vigabatrin and pregabalin demonstrated distinct risk signals for sleep disturbances in our analysis. This finding contrasts with some conventional perspectives suggesting that vigabatrin as adjunctive therapy to carbamazepine does not significantly exacerbate daytime sleepiness (Bonanni et al., 1997). It should be noted that studies supporting this conventional conclusion are often limited by small sample sizes, which may compromise their ability to detect potential risk signals. Therefore, further investigation with higher-level evidence and larger sample sizes is warranted to validate this association. Previous literature suggests that pregabalin may offer potential benefits for sleep in patients with epilepsy, primarily reflected in the improvement of sleep architecture and quality. Some studies indicate that it can enhance sleep depth by increasing slow-wave sleep (deep sleep) and reducing the proportion of light sleep, as well as consolidate sleep maintenance by decreasing the number of nocturnal awakenings (Romigi et al., 2021). Our study failed to replicate these findings. We cautiously speculate that the discrepancy between previous conclusions and our results may be attributable to methodological limitations, along with potential factors such as reporting bias (e.g., misreporting by participants) in the original studies. Cannabidiol exhibits distinct population-dependent effects on sleep. While the drug shows no significant disruption to physiological sleep in healthy populations (Linares et al., 2018), it demonstrates potential for improving sleep microstructure in children with drug-resistant epilepsy. Specifically, Klotz et al. (2021) reported that after 3 months of cannabidiol treatment, over 80% of children with initial abnormalities showed improvements in sleep spindle and slow-wave activity parameters. An open-label study by Anderson et al. (2021) further confirmed that 12 months of cannabidiol treatment significantly improved sleep duration, nocturnal arousals, and daytime sleepiness. However, our analysis identified a signal of disproportionate reporting of sleep disturbances associated with cannabidiol, which contradicts existing research findings. It should be noted that studies supporting the sleep-improving effects of brivaracetam are predominantly open-label and limited in sample size, potentially influenced by expectation bias and population heterogeneity. Therefore, the discrepancies in current conclusions may stem from multiple factors, including study populations, drug dosage, concomitant medications, and research design, necessitating further validation through more rigorous randomized double-blind controlled trials.

Through a comprehensive analysis of ASM-related sAEs, we documented over 30 different sleep-related symptoms, establishing a complete clinical phenotypic spectrum. At the HLGT level, sleep initiation and maintenance disorders were the most common category (51.01%). This distribution pattern is primarily attributed to the preferred term “insomnia” within its subordinate level, with this single symptom alone accounting for 50.4% (1570/3118) of the total reports. This finding aligns with the known properties of ASMs, which have broad neuromodulatory effects and may interfere with sleep-wake pathways (Wang et al., 2018). Notably, such as nightmares, sleep paralysis, and sleep terror, which, although reported relatively infrequently, may profoundly impact patients’ quality of life, mental health, and treatment adherence. Through disproportionality analysis, we detected several potential pharmacovigilance signals (ROR >10). It is crucial to emphasize that these findings represent statistical associations based on a spontaneous reporting system, and their clinical significance requires further validation through additional research. Among these signals, the association between perampanel and sleep disturbances was particularly notable (ROR = 21.42). A similar signal was observed for eslicarbazepine and sleep disturbances (ROR = 20.09). For these drugs, it is recommended that clinicians proactively inquire about changes in patients’ sleep quality and consider enhanced follow-up during the initial treatment phase. Several new findings with clear clinical implications warrant special attention. A significant association was observed between oxcarbazepine and sleep paralysis (ROR = 11.04), a transient state occurring during sleep-wake transitions characterized by voluntary muscle paralysis and vivid hallucinations such as sensed presence or chest pressure, suggesting potential interference with REM-wake transition mechanisms (Denis et al., 2018). Clinicians prescribing oxcarbazepine should proactively inquire about such symptoms, particularly during treatment initiation or dose adjustment, to facilitate early intervention and mitigate impacts on sleep quality and treatment adherence. Additionally, topiramate showed a notable association with Non-24-Hour Sleep-Wake Rhythm Disorder (ROR = 16.25), indicating possible circadian rhythm disruption via mechanisms like carbonic anhydrase inhibition (Dodgson et al., 2000; Parkkila et al., 1995). Patients on long-term topiramate therapy reporting alternating insomnia and excessive daytime sleepiness should be evaluated for circadian rhythm disturbances. Furthermore, adjusting the dosing schedule may serve as a management strategy for these individuals. Furthermore, levetiracetam demonstrated a strong association with increased sleep need (ROR = 20.06), underscoring the importance of pre-treatment counseling about daytime sleepiness, assessing its functional impact, and warning against high-alertness activities during therapy. These signals provide valuable references for clinical practice, but their causal relationships and clinical significance need further confirmation through prospective studies and larger samples. Future research should focus on elucidating the underlying mechanisms of these associations and providing clinicians with evidence-based management guidelines.

Time-to-event analysis revealed differences in the timing of drug-related signals; however, interpretation must account for potential issues such as recall bias, missing data, and left truncation. In this study, the proportion of cases with complete onset data ranged from 6.8% (gabapentin) to 32.4% (perampanel), with most drugs below 25%, indicating that incomplete data are common. Against this background, drugs such as eslicarbazepine (9.6%), pregabalin (10.1%), and zonisamide (25.7%) exhibited early disproportional reporting patterns, with a median time to onset of approximately 10 days, suggesting a possible association with rapid pharmacokinetic attainment or acute pharmacodynamic effects. In contrast, phenytoin (14.3%) demonstrated a markedly delayed reporting pattern, with a median onset time exceeding 4 years, which was significantly longer than the 1–3 months observed for most other drugs. It should be noted that date entries in spontaneous reporting systems are often incomplete or inconsistent. The exceptionally delayed onset observed with phenytoin may particularly reflect reporting artifacts, such as recall bias, rather than a true delayed pharmacodynamic effect. Despite these limitations, the observed temporal patterns still suggest the value of individualized monitoring strategies, such as intensified follow-up for early-signal drugs and long-term monitoring mechanisms for delayed-signal drugs; however, all such strategies must be developed with full awareness of the inherent uncertainties in spontaneous reporting data.

Several limitations of this study should be acknowledged. First, disproportionality analysis can only indicate statistical associations rather than establish causal relationships. Second, the FAERS database, as a spontaneous reporting system, is subject to several inherent biases. These include selective reporting bias, potential underreporting of certain events, overreporting influenced by media attention or regulatory actions, and channeling bias (where drugs are preferentially prescribed to patients with specific characteristics). Additionally, the absence of comprehensive denominator data precludes estimation of true incidence or prevalence rates, and thus findings should not be interpreted as reflecting actual population risks. Important clinical confounders such as disease severity, specific ASM dosage, treatment duration, and detailed patient comorbidities are not routinely available, which could influence the observed associations. Furthermore, the wide confidence intervals for rare events reflect inherent estimation uncertainty.