Sleep Patterns and Issues in Lennox-Gastaut Syndrome: An Overview

Studies were included if they enrolled individuals with a diagnosis of LGS, regardless of age or etiology, or mixed complex epilepsy cohorts in which at least 50% of participants had LGS. We did not exclude individuals based on the presence of other comorbid conditions. Eligible studies assessed sleep-related parameters, including macrostructural measures (e.g., distribution of rapid eye movement (REM) and non-rapid eye movement (NREM) stages, total sleep time [TST], total time in bed [TIB], sleep efficiency [ratio of TST to TIB], sleep latency [time from lights-out to sleep onset], wake after sleep onset [WASO], and REM latency), microstructural features (e.g., cyclic alternating pattern [CAP]), nocturnal seizure patterns, and interictal epileptiform discharges (IEDs) in relation to sleep cycles, as well as sleep-disordered breathing (SDB) or circadian rhythm alterations.

Sleep architecture (macrostructure) refers to the organization of NREM (N1–N3) and REM stages across sleep cycles, while sleep microstructure examines finer oscillatory dynamics reflecting brain stability, arousal regulation, and sleep quality [8]. A key microstructural feature of NREM sleep is the Cyclic Alternating Pattern (CAP), which alternates between Phase A (transient activation) and Phase B (stable NREM), recurring every 20–60 s [9]. Phase A is subclassified into A1 (stable slow waves), A2 (mixed slow and fast components), and A3 (desynchronized fast activity), with the CAP rate—the proportion of NREM occupied by CAP—indicating sleep fragmentation. Predominant A1 phases reflect restorative sleep, whereas higher A2/A3 activity indicates disrupted sleep. REM density, measured from electrooculography signals, represents the frequency of rapid eye movements per unit time during REM and reflects REM sleep intensity.

Apneas and hypopneas were scored according to the updated American Academy of Sleep Medicine (AASM) 2012 criteria [10]. In adults, an apnea was defined as a ≥90% drop in airflow from pre-event baseline, measured by an oronasal thermal sensor, positive airway pressure (PAP) device flow, or alternative sensor, lasting ≥10 s. Apneas can be further classified as obstructive, in which airflow ceases due to upper airway obstruction despite ongoing respiratory effort, or central, in which airflow stops due to absent respiratory effort, often related to impaired brainstem control. Hypopneas were defined as a ≥30% drop in airflow for ≥10 s, associated with either ≥3% oxygen desaturation or an arousal. Classification of hypopneas as obstructive or central in adults is optional. In children, apneas were defined as a ≥90% drop in airflow for ≥2 breaths and classified as obstructive, central, or mixed. Central apneas in children required absence of inspiratory effort plus at least one of the following: duration ≥20 s, ≥3% oxygen desaturation or arousal, or (for infants <1 year) a decrease in heart rate below defined thresholds. Hypopneas in children were defined as a ≥30% drop in airflow for ≥2 breaths, associated with ≥3% oxygen desaturation or an arousal. Surrogates of arterial PCO₂ (end-tidal or transcutaneous) were used when applicable. The apnea–hypopnea index (AHI) quantifies the average number of apneas and hypopneas per hour of sleep and is commonly used to assess the severity of sleep-disordered breathing. Because several included studies were conducted prior to the AASM 2012 [10] update, apnea and hypopnea events were extracted from each study based on the scoring methods reported in the original publications.

Included study designs comprised observational studies (cross-sectional, case–control, or cohort), interventional trials, and case series with ≥3 participants, conducted in clinical, neurophysiological, or interventional settings using polysomnography (PSG), video-EEG, actigraphy, or sleep questionnaires. Exclusion criteria were animal studies, single case reports, reviews, editorials, or studies not reporting sleep-specific outcomes in LGS, as well as those describing only nocturnal tonic seizures captured on EEG or PSG without sleep stage–specific information. Only studies published in English were considered.

A comprehensive search of PubMed, Embase, Ovid, and ClinicalTrials.gov was conducted from database inception to October 2025. The search combined controlled vocabulary and free-text terms related to LGS and sleep. The primary search string, adapted for each database, was: ("Lennox-Gastaut" OR "Lennox Gastaut" OR "epileptic encephalopathy") AND ("sleep" OR "polysomnography" OR "REM" OR "NREM" OR "sleep architecture" OR "cyclic alternating pattern" OR "sleep disorder" OR "sleep apnea" OR "circadian" OR "melatonin"). Additional studies were identified through reference lists of included articles and relevant reviews.

All citations were imported into EndNote for duplicate removal and screening. Titles and abstracts were reviewed for relevance, followed by full-text assessment against eligibility criteria, with reasons for exclusion documented at the full-text stage. Due to heterogeneity in study designs and outcome measures, quantitative meta-analysis was not feasible; findings were synthesized narratively and organized thematically across domains. Methodological rigor and reporting quality were assessed using the Joanna Briggs Institute (JBI, Adelaide, South Australia, Australia) checklist for analytical cross-sectional studies, case–control, case series, and cohort studies, as appropriate.

After screening 1242 articles, eleven studies met inclusion criteria, comprising nine full-length publications, one conference abstract, and one registered but unpublished clinical trial (Figure 1). The studies, conducted between 1986 and 2025, were carried out across Europe, Asia, Australia, and North America. (Table 1) Most were single-center observational studies, including six retrospective and three prospective designs. Sample sizes were small, ranging from 5 to 16 patients with LGS, with or without control groups (epilepsy or healthy populations). Two studies included mixed DEE cohorts with more than 50% of participants having LGS. Two studies focused specifically on adults with LGS, while the remaining studies included pediatric or mixed-age cohorts.

Polysomnography (PSG) was the primary assessment method in eight studies, with one study using ambulatory wearable monitoring and another assessing circadian biomarkers (melatonin, cortisol, and body temperature). The most common objectives were characterization of sleep architecture, microstructural correlates of epileptiform discharges, and identification of respiratory or circadian abnormalities.

Across studies, patients with LGS consistently exhibited marked alterations in sleep architecture compared with healthy or epilepsy controls. In one study, REM sleep was either absent or profoundly reduced in 6 out of 11 children with LGS, with total REM proportion significantly lower than in both epilepsy controls and healthy subjects [11]. These findings were replicated in another study, where REM sleep duration was reduced by more than half compared to controls (55 vs. 116 min), accompanied by a compensatory increase in slow-wave sleep [13]. The total sleep time, wake after sleep onset, and sleep latency did not differ significantly. In contrast, another study reported reductions in time in bed, total sleep time, and sleep efficiency, along with increased REM latency and a higher number of stage shifts per hour [12]. The study by Horita et al. in younger children similarly demonstrated reduced REM percentage and REM density [15]. Sleep spindles were often absent or markedly attenuated in 60%, particularly among children with severe intellectual disability, highlighting the difficulty in NREM sleep substaging in some of these patients.

In a study of 10 patients with LGS undergoing evaluation with implanted electrodes in the centromedian (CM) thalamus, patients exhibited longer wakefulness and shorter stage II sleep compared with healthy controls. Among patients with LGS, those experiencing seizures had significantly more REM periods with shorter latency than patients without seizures [19]. Another study by the same group, involving five patients with LGS, employed all-night EEG recordings using thalamic depth electrodes alongside concurrent scalp EEG to characterize both normal and abnormal stage II slow-wave sleep (SWS II) patterns [20]. A total of 1439 SWS II events (1233 normal and 206 abnormal) were analyzed using visual and statistical temporo-spatial correlation methods [20]. Most SWS II activities were concurrently observed in both scalp and CM recordings. Normal spindles, typically generated through CM–cortical interactions, were larger when synchronized and modulated other thalamo-cortical events [20]. In contrast, abnormal spindles originated from widespread cortical and CM regions, disrupting physiological rhythms. Vertex waves localized to parietal scalp regions, and overall findings implicated thalamo-cortical circuits in the generation of both normal and abnormal SWS II activities [20].

Sleep microarchitecture has been examined in limited studies, most notably by Eisensehr et al., who analyzed the CAP, a marker of NREM instability [13]. Patients with LGS exhibited a markedly elevated CAP rate (68% vs. 33% in controls), and generalized polyspike bursts were significantly more frequent during CAP phases, particularly phase A [13]. The frequency of generalized polyspike bursts during NREM sleep correlated with both the number of A phases containing polyspikes and their mean polyspike count [13]. In adults, Sforza et al. confirmed that interictal epileptiform discharges (IEDs) were sleep stage–dependent, peaking during NREM stage 2 and slow-wave sleep, and lowest in REM [17]. The IED rate was highest during the first three hours of the night, suggesting a circadian or homeostatic influence on epileptic activity. Another study also reported more frequent interictal and ictal discharges during NREM sleep [15]. A study using thalamic depth electrodes with concurrent scalp EEG demonstrated that seizure occurrence and the duration of the late seizure component were markedly increased during stage II sleep and decreased during REM sleep. Interictal spike–wave discharges were significantly more prominent during NREM sleep and attenuated during REM sleep compared with wakefulness [19].

Sleep-disordered breathing (SDB) in LGS has received limited attention but is increasingly recognized as an important comorbidity. SDB in LGS encompasses both apneas and hypopneas, which are disruptions in airflow during sleep.

In a cohort of children with DEEs, including 14 with LGS (n = 23), patients exhibited significantly worse respiratory parameters—AHI/hour, oxygen desaturation index/hour, mean peripheral oxygen saturation (SpO₂), and SpO₂ nadir (all p < 0.001)—as well as higher periodic limb movements (PLMs%, p < 0.001) compared with 40 healthy controls [12]. In another study of adults with DEEs, including six with LGS, full polysomnography with respiratory monitoring revealed that over half of participants had moderate-to-severe obstructive sleep apnea (OSA; AHI ≥ 15), and nearly one-third had severe OSA (AHI ≥ 30) [18]. Tonic seizures were frequently associated with central apneas, sometimes representing their only clinical manifestation. Contributing factors included antiseizure medication–related weight gain, benzodiazepine use, and structural craniofacial abnormalities.

One study investigated circadian organization in 16 institutionalized individuals with LGS aged 8–45 years by measuring salivary melatonin, cortisol, and axillary temperature every two hours over 26 h [16]. Of the 9 subjects with normal sleep–wake patterns (group 1), 2 showed abnormalities in one or more rhythms, whereas 6 of the 7 subjects with disordered sleep (group 2) had rhythm disruptions. All three rhythms were disrupted in 2 subjects from group 2, who were also the only ones with abnormal cortisol rhythms, highlighting a higher prevalence and severity of circadian dysregulation in those with sleep disorders.

Recent work has explored more accessible sleep assessment methods. A 2025 pilot feasibility study demonstrated that home-based wearable devices, such as the Apple Watch (consumer-grade device that tracks sleep, heart rate, and activity via photoplethysmography and accelerometry), and WatchPAT (a wrist-worn home sleep apnea testing device indicated in individuals 12 years and older) that estimates sleep stages and respiratory indices using peripheral arterial tone, heart rate, and oxygen saturation) were acceptable and practical for children with LGS, enabling multiweek monitoring outside the hospital setting [14]. The study reported an average monitoring duration of approximately 16.9 days and demonstrated that the use of wearable sleep monitoring devices was feasible in LGS. Families generally favored wearable devices over traditional modalities. Among various assessment methods, parents rated questionnaires as the most convenient (mean score 4.1 ± 0.9), followed by the Apple Watch (3.7 ± 1.6), whereas WatchPAT and polysomnography were rated as least convenient (2.4 ± 2.1 and 2.4 ± 2.0, respectively). The Apple Watch was the most preferred tool overall (1.6 ± 1.0), highlighting its practicality and user comfort in this population. The WatchPAT was moderately preferred (3.3 ± 1.4), ranking above polysomnography. In contrast, polysomnography remained the least preferred method (4.1 ± 1.4). Comparison of apnea–hypopnea index (AHI) and peripheral respiratory disturbance index (pRDI; a metric used in home sleep testing devices like WatchPAT) values showed concordance between two of the three sleep apnea tests, with one yielding discrepant results.

A registered but unpublished randomized, double-blind, crossover trial (NCT01370486) aimed to evaluate the effect of melatonin on sleep architecture and epileptiform discharges in LGS. The protocol specified PSG assessments before and after treatment, with primary outcomes including ≥50% reduction in nocturnal discharges and ≥15% increase in slow-wave sleep. However, the trial was subsequently withdrawn, and no published results are currently available.

This scoping review has several limitations. The primary studies were small, single-center investigations, all of which exhibited significant selection and other biases, including convenience sampling, often unmatched controls, lack of blinding in scoring, single-night assessments, and inadequate adjustment for age, etiology, or medications. Cohorts are heterogeneous with respect to age, etiology, medication regimens, and recording environments (inpatient versus ambulatory). Key factors affecting sleep and respiration—such as medication use, neuromodulation status, and comorbidities like obesity or craniofacial anomalies—are inconsistently reported. Although LGS diagnoses were consistently reported in the primary studies, they may not fully align with the current International League Against Epilepsy (ILAE) 2022 criteria due to evolving definitions [1]. The ILAE 2022 criteria are more specific than earlier definitions, focusing on multiple seizure types, characteristic EEG patterns, and cognitive impairment [59]. Some patients previously diagnosed with LGS may not meet current criteria, while others may now be classified under broader DEE categories. The impact of these changes on the interpretation of older study findings is uncertain. Many studies emphasize electrophysiologic outcomes (e.g., interictal discharge frequency, CAP indices) rather than patient-centered measures such as daytime functioning, cognition, or quality of life. Additionally, inpatient polysomnography may not reflect habitual sleep, and seizure-related apneic events shorter than 10 s are often unrecognized by standard sleep-disordered breathing scoring, despite their potential clinical relevance. Evidence regarding the relationship between tonic seizures and central sleep disturbances in LGS is very limited, with only one study reporting an association between frequent tonic seizures and central apneas, highlighting a need for further investigation. Finally, parasomnias have not been systematically studied or clearly characterized in patients with LGS, representing a knowledge gap and a potential challenge in differentiating nocturnal seizures from parasomnia-like behaviors.

Clinicians should maintain a high index of suspicion for sleep disturbances in LGS. (Table 2) Screening for sleep-disordered breathing using questionnaires and, when feasible, polysomnography is appropriate—particularly for adults and children with risk factors such as obesity, craniofacial abnormalities, chronic benzodiazepine use, or VNS therapy. Sleep assessment during routine EEG or overnight video-EEG can also be valuable, as careful visual analysis of the sleep portions may provide important insights. Medication choices and dosing schedules should take into account effects on REM sleep and respiratory drive. When SDB is identified, standard interventions such as CPAP, weight management, or ENT evaluation may improve not only sleep but also seizure control and daytime functioning, although direct evidence in LGS remains limited.

Future studies should pursue large, multicenter, prospective polysomnographic cohorts using standardized sleep scoring, full EEG montages, and harmonized reporting of ASMs, neuromodulation, body mass index (BMI), and etiology, encompassing both pediatric and adult patients to define age-related trends [60]. Ambulatory and longitudinal sleep monitoring—combining validated wearables with sleep diaries—could capture habitual sleep, seizure–sleep interactions, and long-term outcomes. Quantitative EEG and CAP analyses should be integrated with seizure timing to determine whether pharmacologic or behavioral CAP modulation reduces nocturnal epileptiform activity. Interventional trials assessing whether SDB treatment (CPAP, positional therapy, weight loss) or circadian-targeted interventions (melatonin, light therapy) improve seizures, cognition, and quality of life are urgently needed. Mechanistic studies using intracranial EEG or neuroimaging should dissect thalamocortical contributions to CAP and generalized discharges, potentially identifying novel therapeutic targets (e.g., REM/NREM modulation or orexinergic signaling). Future studies should investigate how specific genetic variants associated with LGS may contribute to distinct sleep phenotypes, which could inform personalized management strategies.

Emerging work also underscores the cognitive relevance of sleep stages: slow-wave sleep supports hippocampal–neocortical consolidation of declarative memories, while REM sleep promotes emotional memory integration and synaptic plasticity [61,62]. Disruptions in either stage correlate with cognitive decline across neurologic disorders. Future studies in LGS should test whether targeted neurostimulation—such as closed-loop slow-oscillation stimulation, thalamic or hippocampal DBS, or responsive neurostimulation (RNS) with phase-locked delivery—can enhance circadian changes in aperiodic activity, slow wave sleep, or REM architecture and improve cognitive resilience in LGS [63].