2.1 Objectives

The primary objective was to describe data on the burden of illness of LGS globally, across the domains of clinical symptom burden, comorbidities, care requirements, humanistic burden on individuals and caregivers, economic burden, and treatment burden. The secondary objective was to identify reasons for variations in findings across the included studies, and any current data gaps.

2.2 Outcomes

The main outcomes were clinical symptom burden (types and patterns of seizure and nonseizure symptoms), prevalence of comorbidities, care requirements, humanistic burden on individuals and caregivers, economic burden, and treatment burden.

2.3 Eligibility criteria

Noninterventional studies of any design were eligible for inclusion if they presented original research on one or more outcomes of interest and included primary or subgroup analyses in an LGS population (Table S1–S6). Inclusion was not restricted by population, country of origin, or publication date. Publications written in English and non-English publications with an English abstract were considered for inclusion. Preclinical studies, case reports, reviews, and interventional studies in which individuals were actively assigned to a therapy were excluded. For this review, the use of screening, diagnostic tests, or assessments (including qualitative interviews) was not defined as intervention.

2.4 Data sources

Relevant literature was identified via MEDLINE, Embase, and the Cochrane Library from inception to March 15, 2022. For the period 2017–2022, supplementary manual searches were also performed by one reviewer (E.B.) to identify relevant proceedings from key congresses, websites, and publications listed in bibliographies of review articles (Table S2).

2.5 Search strategy

Two reviewers (A.J and E.B.) used a combination of Medical Subject Headings and free-text terms to identify relevant literature (Table S2). The results from MEDLINE, Embase, and Cochrane Library searches were downloaded via Endnote into a bespoke Microsoft Excel database. After deduplication, titles and abstracts, followed by full texts, were screened in duplicate by independent reviewers (E.B., A.J., and nonauthor reviewers). Discrepancies were resolved by reviewer discussion or escalation to a third independent reviewer (E.B., A.J., and nonauthor reviewers). Articles identified during supplementary manual searches were added to the final included list of publications after full-text screening.

2.6 Data extraction and quality assessment

Data on the study design, population, and outcomes were extracted from publications by a single reviewer for each article. This reviewer also assessed study quality based on a series of “risk of bias” questions adapted from the Strengthening the Reporting of Observational Studies in Epidemiology statement,¹³ the Appraisal Tool for Cross-Sectional Studies,¹⁴ and the Newcastle–Ottawa quality assessment scales.¹⁵ Risk of bias was rated as high, medium, or low across eight categories relating to study design, methodology, and reporting, to reflect potential concerns over the reliability and generalizability of the reported results. Seizure types were recorded exactly as they were described in each article, which included seizure type terminology that has since been discontinued.¹⁶

2.7 Data synthesis

Owing to heterogeneity and the qualitative nature of some outcomes, no quantitative synthesis or subgroup analyses were performed. Results are presented as narrative and tabulated summaries, with descriptions of themes, variation, and quality concerns. Costs describing the economic burden were standardized to a single currency, when possible, accounting for or noting potential differences by year of publication, inflation, and other economic conditions.

3.1 General findings

After screening 1442 deduplicated records from databases and registers, 108 were eligible for inclusion in this review (Figure 1). Five additional articles were identified in the supplementary searches, leading to a total of 113 included articles (Figure 1). Most of the included studies reported on the clinical burden of seizure symptoms, treatment burden outcomes, or both (Figure 2). Conversely, there were few reports on caregiver burden or humanistic burden.

North America had the highest representation of studies (n = 39), followed by East Asia (n = 32) and Europe (n = 31) (Table S3). Other regions, including South America (n = 9), western Asia (n = 7), Oceania (n = 3), and Africa (n = 2), were poorly represented.

3.2 Quality assessment findings

A summary of the qualitative assessment results is given in Table S4. Briefly, few quality issues were noted in terms of study design, methodology concerns, and clarity of reporting in the Results and Discussion sections. Quality concerns were noted for sample representativeness (n = 95), disease and outcome measures used (n = 57), and overall reporting clarity (n = 39). These concerns were raised owing to the inclusion of small or selective samples, a lack of clarity on or confirmation of LGS diagnosis, unclear selection criteria, and limited reporting clarity because of the type of publication (i.e., abstract). Although no studies were excluded owing to quality assessment findings, quality concerns were considered when analyzing the extracted data and reporting findings.

3.3 Clinical symptom burden and comorbidities

3.3.1 Seizures

Of the 72 studies reporting the burden of clinical seizure symptoms, six studies (two each in Asia, Europe, and North America) included the age at first seizure, which ranged from 1 month to >9 years. No clear differences by geographical location or publication date were apparent.^17-22 Studies including both children and adults (n = 3) reported an average age at first seizure of 1–3 years,^20-22 which is older than that reported in the studies that included children only (n = 3).^17-19 Across eight studies with relevant data, 0%–60% of individuals with LGS experienced infantile spasms, either before the onset of epilepsy or as the first recorded seizure type.^19-26

Data from 17 studies showed that individuals with LGS generally experience 2–4 concurrent seizure types, although substantial interindividual heterogeneity existed, with an overall range of 0–7 seizure types per individual.^{20, 21, 26-40} Five studies reported the mean number of seizure types per individual, which was between 2.6 and 3.6.^{28, 37-40} These data were generally aligned with those from studies that reported the proportion of individuals experiencing different numbers of seizure types (Figure 3).^{20, 21, 26-36} One study reported a high prevalence of individuals with one seizure type (38%); however, this study used historical records of seizure types experienced over 14 years.²⁹

Four studies reported a decrease in the number of seizure types experienced per individual during the study, which may reflect changes with age.^{20, 27, 34, 41} Similarly, a decline in the prevalence of individual seizure types (tonic, tonic–clonic, myoclonic, and atypical absence) over time was reported in multiple studies (Table S5).^{20, 36, 38, 42, 43} Three studies also noted reductions in the prevalence of atonic seizures over time.^{20, 36, 43}

In all 11 studies that reported continuous measures of seizure frequency, individuals with LGS were found to experience more than one seizure per day.^{29, 30, 37, 41, 44-50} Seizure frequency ranged widely within and across studies, which may be partly owing to interindividual heterogeneity; in one study, baseline seizure frequency was 15–3000 seizures per month.^{46, 51} These findings aligned with those from studies in which individuals were categorized by seizure frequency; 50%–100% of individuals were reported to have daily seizures, and those without daily seizures were usually reported to have seizures at least once per week.^{18, 20, 23, 26, 28, 34-36, 52-58}

Several prospective and retrospective studies showed decreases in seizure frequency over time, reporting similar ranges of seizure freedom prevalence at last visit (9%–28% and 4%–28%, respectively).^{17, 20, 34, 38, 43, 47, 59-67} Although the duration of data collection varied considerably between studies (3–20 years), no correlation between the change in seizure frequency and the duration of data collection was observed.^{20, 34, 38, 43, 47, 59-64} One study showed that 33.3% of children treated with clobazam had seizure-free status maintained for a period of 5.5 months, 25% of children treated with topiramate had seizure-free periods ranging from 8 months to 2.5 years, and 11% of children treated with lamotrigine had seizure-free periods of 6 months.²⁶ In another study, 70.6% of children had at least one seizure-free period up to their fifth birthday, with a median longest seizure-free period of 13.5 months.¹⁷

The prevalence and frequency of seizure types observed in individuals with LGS are summarized in Table S5. Tonic seizures showed the highest prevalence of all reported seizure types, followed by tonic–clonic and myoclonic seizures. Several definitions were used for atonic, astatic (now referred to as atonic),¹⁶ and absence seizures, and status epilepticus, precluding a clear conclusion on prevalence. Additionally, some studies reported the prevalence of drop attacks, which can encompass several seizure types, including atonic, tonic, and generalized tonic–clonic seizures.

3.3.2 Nonseizure symptoms and comorbidities

Overall, 59 studies reported on nonseizure symptoms or comorbidities. In the present review, all data are combined and original study terminology (i.e., “nonseizure symptom” or “comorbidity”) is used when relevant. Distinguishing between nonseizure symptoms and comorbidities was difficult given the overlapping definitions across studies, with 47 studies reporting on nonseizure symptoms (Table S6) and 12 studies reporting on comorbidities (Figure 4). Overall, the most common nonseizure symptoms or comorbidities were developmental delay, intellectual disability, and motor impairments.

Developmental delay was common in individuals with LGS, with eight studies reporting a prevalence of 79.6%–100% (Table S6). These delays were often severe or profound (>50% prevalence based on three articles)^{21, 25, 68} and associated with developmental regression (44% prevalence based on one article).⁶⁵ Furthermore, based on 19 studies, at least 90% of individuals with LGS had intellectual disability, often of moderate to profound severity (reported by 14 articles to be present in at least 50% of the sample), and four studies noted that severity may worsen over time. Compared with studies in which it was described as a nonseizure symptom, the prevalence of intellectual disability when described as a comorbidity was more varied (59% and 100% in two studies).^{33, 69}

Eleven studies reported on motor impairments, which were frequent in individuals with LGS, with one study reporting difficulty walking in 81% of individuals.⁷⁰ However, variation in the measures and definitions used prevented definitive conclusions on the prevalence and severity of specific motor problems (e.g., motor handicap, walking/sitting) or symptoms (e.g., ataxia, hypotonia, paresis, spasticity).

Based on six studies, many individuals with LGS showed difficulties with communication, with up to 60% of individuals reported to have speech disorders (including dysarthria) and one study showing 60% of patients to be nonverbal.^{21, 26, 29, 68, 70, 71} In one study (n = 24 parents), 96% of parents reported that their child could not communicate appropriately for their age.⁷⁰ Two studies described comorbidities of language disorders, reporting a prevalence of 40% and 42%.^{33, 72}

Autism^{33, 36, 58, 60, 69, 73} and cerebral palsy^{25, 32, 72-74} were comorbidities reported in six and five studies, respectively, with prevalence estimates of 6%–13% and 22%–33% of individuals, respectively (Figure 4). The prevalence estimate for autism excluded studies with unreliable estimates, namely one study with only five individuals³³ and another study in which “probable” LGS only was identified.⁶⁹ Although the latter study described patients with “probable” LGS and was excluded from the prevalence estimate for autism, the article did meet criteria for inclusion in the present review, because patients were identified from insurance claims records using a machine-learning model based on treatment- and diagnosis-related data. Identified patients were also reviewed by clinicians to ensure suitability for inclusion. In the present review, the estimate for cerebral palsy excluded studies with small or selective samples, namely a study of vagus nerve stimulation (VNS)⁷³ and another with a small sample size (23 individuals).²⁵ If included, these studies would change the range of prevalence estimates for autism and cerebral palsy to 0%–20% and 22%–56%, respectively.

Other symptoms were grouped into comorbidities defined as psychiatric and behavioral issues, and were reported in five studies^{24, 26, 68, 75, 76}; however, methods and findings were heterogeneous, preventing any clear conclusions regarding prevalence. One study reported that 76% of individuals had behavioral issues⁶⁸; another study found that 80% of individuals had hyperactivity, aggressiveness, and autistic features.⁷⁵ Conversely, other studies reported lower prevalence, potentially owing to their focus on specific symptoms (e.g., hyperactivity)⁷⁵ or behavioral syndromes (e.g., anxiety, attention-deficit/hyperactivity disorder, or psychosis).⁷⁶ Studies that described psychiatric and behavioral issues as comorbidities generally reported lower prevalence (35%–42%) than those that described them as nonseizure symptoms (35%–50%).^{36, 72, 77}

Other problems included difficulty swallowing (reported in two studies)^{43, 70} and strabismus (reported in one study with a prevalence of 5.3%).³⁶ Based on five studies, individuals with LGS spent less time in sleep stage 2 and in rapid eye movement sleep than those without LGS.^{31, 68, 78-80}

Prevalence estimates of comorbidities reported in only one study each were overweight/obesity (80%)⁸¹; acute upper respiratory tract infections (44%)⁷²; bone fracture (31%)⁶⁹; postviral encephalitis (6%)⁷⁴; insomnia (5%)³⁶; Down syndrome (4%)⁷⁴; hypothyroidism (3%)³²; myelomeningocele (2%)⁷⁴; cancer (2%)³²; diabetes, hypopituitarism, and renal failure (1% each)³²; and urinary incontinence, osteoporosis, and gastrointestinal conditions (not specified).³⁶

3.4 Care requirements

Overall, 24 studies assessed the care requirements of individuals with LGS, which were found to be considerable owing to the high prevalence of seizure and nonseizure symptoms, as well as frequent comorbidity. In terms of daily assistance, 60.3% of adults (n = 41) in one study lacked independent daily living skills, including bathing, eating, functional mobility, and toileting.⁸² Thirteen studies reported that individuals with LGS frequently have problems with mobility and eating, including difficulty walking, lack of functional grasp strength, and a requirement for tube feeding (Table 1).^{10, 21, 26, 28, 45, 49, 58, 68-70, 82-84}

The considerable clinical symptom burden associated with LGS means that individuals often require education outside of mainstream settings,^{30, 59} and few individuals complete high school (5.1% and 22.6% completion rate in two studies) or undertake higher level education (Table 1).^{22, 32, 38} Additionally, many individuals cannot live independently or support themselves economically, often requiring full-time caregiver support in childhood and residential care in adulthood (Table 1).^{27, 30, 36, 38, 42, 45, 49, 77, 85}

3.5 Humanistic and caregiver burdens

Only three studies provided information on the QoL of individuals with LGS, with sample sizes ranging from 40 to 416.^{9, 68, 86} Given the absence of validated scales for assessing the QoL of individuals with LGS, this outcome was not measured directly in any of the studies; instead, assessment was reliant on caregiver interviews or surveys. Moreover, no studies directly questioned individuals with LGS, probably owing to the intellectual disability associated with LGS. In one study, a conceptual model was developed to reflect four identified themes around the impact of LGS on QoL (physical, cognitive and behavioral, social, and treatment).⁹ These themes were consistent with findings from two caregiver surveys, one of which included carers of individuals with other epilepsies (Figure 5).^{2, 6} Specific aspects of the identified themes could have direct or indirect effects on QoL (e.g., developmental delay has a direct impact and affects the ability to make friends), and certain aspects may have a positive or negative impact, such as treatments (depending on efficacy).^{9, 68}

In one study, individuals and caregivers (for Dravet syndrome, LGS, or other epilepsies) provided expected patient QoL scores based on a series of hypothetical vignettes. A strong positive correlation was reported between hypothesized patient QoL scores and the number of seizure-free days per month (p < .001). Another study assessed patient QoL using a mean health utility score, obtained by converting visual analog scale scores to a 0–1 scale by dividing by 100. This study showed that seizure frequency greatly affected patient QoL, with the lowest reported mean health utility score (.14) being associated with the highest frequency of drop seizures (130 seizures and one seizure-free day per month).⁸⁶

Five studies interviewed parents and caregivers to assess the burden they personally experienced, with multiple countries represented, including Australia, France, Italy, the UK, and the USA.^{9, 10, 86-88} Effects on caregivers were noted in four domains: physical, social, emotional, and work/finances (Figure 5).^{9, 87}

The reported emotional burden on caregivers was multifaceted; caregivers often went through stages of adjustment after learning that their child had LGS, beginning with disbelief, bargaining, fear, and anger, and followed by acceptance and identity readjustment.¹⁰ Although this burden affects all family members, the greatest burden is experienced by the main caregiver, who is usually the mother of the affected child.⁸⁷ Caregivers experience uncertainty around the diagnosis and cause of LGS, seizure activity, treatment efficacy, and prognosis, but parent support groups can reduce isolation and worry, and provide parents with practical advice.⁸⁸

3.6 Economic burden

Overall, 13 articles reported on economic burden, with nine studies from the USA and one study each from Germany, Japan, Mexico, and the UK. Of these 13 articles, 11 included information on costs and seven included information on health care usage, with five articles providing information on both (Figure S1). The articles on costs generally focused on direct costs, with little or no information on indirect costs such as lost caregiver productivity and days lost from work.

Direct costs varied considerably across studies, even for studies in the same country. Generally, USA-based studies reported higher per-patient-year costs ($29 911–$80 545) than studies in other countries, including Germany ($24 450.91) and Mexico ($707.34–$2123.25). These costs were often categorized as medical service or pharmacy costs. Medical service costs were generally higher than pharmacy costs; however, there was considerable variation. The main contributor to medical service costs was inpatient costs, which composed 21%–60% of the mean annual health care cost.^{67, 72, 89}

One USA-based study reported a mean (SD) cost of $8147 ($43 218) per medically treated seizure event for Medicaid-insured individuals and $14 759 ($43 600) for commercially insured individuals.⁶⁷ Another USA-based study showed that costs for LGS were approximately three times higher than for other epilepsies.^{69, 90} Pharmacy costs ($1592–$24 018) were mainly driven by the high costs of ASMs, with some evidence of rising costs over time.^{69, 90-92}

Across five studies, inpatient admissions ranged from .6 to 3.6 per patient-year,^{67, 69, 72, 89, 93} with a mean length of each hospital stay of 2–5 days.^{72, 93} Four studies showed that outpatient visits ranged from 6.3 to 11.8 per patient-year.^{69, 84, 89, 94}

3.7 Treatment burden

Most individuals with LGS receive ASMs to manage their symptoms.^{23, 26, 28, 30, 34, 36, 42, 53, 55, 60, 69, 78, 80, 82, 83, 88, 91, 95-100} Across 23 studies with information on LGS treatment, 19 reported that all individuals with LGS received ASMs, and the other four studies generally found that more than 80% of individuals used ASMs. However, pharmacoresistance associated with switching of ASMs^{21, 23, 34, 36, 39, 41, 45, 49, 50, 72, 73, 77, 85, 92, 93, 95, 96, 101-104} and polytherapy^{51, 105} was common. These findings were consistent for children and adults across several countries, suggesting no difference by demographic or individual characteristics.

Across four studies, 73%–100% of individuals met study-defined criteria for pharmacoresistance.^{101, 105-107} These four studies covered relatively small samples of children from Italy, Nepal, South Korea, or the USA, but did suggest pharmacoresistance in LGS is not limited to a particular pediatric age group or geographical location. Polytherapy is usually required,^{26, 30, 60, 97, 105} with concurrent use of 2–4 ASMs per individual with LGS reported across approximately 40 studies.^{18, 28, 34, 36-39, 42, 44, 48-51, 53-55, 58, 66, 68, 72, 73, 78, 80, 81, 83, 88, 90, 95, 96, 98-100, 102, 104, 108-111} Exceptions to the finding that 2–4 ASMs are used concurrently by individuals with LGS were reported in some studies. For example, a UK-based study in 256 individuals with LGS (110 confirmed, 146 probable) estimated a lower mean number of ASMs used per year (approximately one).⁹³ However, switching of ASMs was common, and the mean total number of ASMs used by individuals with confirmed or probable LGS was 6.7 (SD = 3.4), and 8.5 (SD = 3.5) over a mean follow-up time (averaged across confirmed and probable LGS) of 11.7 (SD = 8.2) years.⁹³ Six studies reported that individuals with LGS generally used more than four ASMs concurrently,^{65, 69, 84, 85, 112, 113} including one retrospective analysis from the USA (n = 14 712), which used data from the Medicaid multistate database and reported a mean of 5.8 (SD = 2.3) concurrent ASMs after a mean observation period of 11.1 (SD = 4.5) years.⁶⁹ These data suggest that use of ASMs can vary widely across individuals with LGS.^{65, 69, 84, 85, 112, 113}

The ASM used by the highest proportion of individuals was valproate or valproic acid (30.8%–100%), with upper estimates for the frequency-of-use range being ≤80% for all other ASMs (Table S7).^{23, 28, 30, 34, 45, 48, 49, 53, 56, 58, 66, 73, 77, 78, 82, 85, 93, 95, 98-100, 108-110, 114} Reported treatment combinations varied, but common combinations across four studies with data were valproate or valproic acid with lamotrigine, clobazam, or levetiracetam.^{66, 72, 74, 99} Medications reported by fewer than four studies included drugs approved by the US Food and Drug Administration (FDA) and the European Medicines Agency for use in treatment of epilepsy decades ago (acetazolamide, ethosuximide, mesuximide, oxcarbazepine, phenytoin, and primidone), those with more recent FDA approval (lacosamide and midazolam), and those used off-label to treat epilepsy (cannabis-based medicinal products, intravenous gamma globulin, potassium bromide, and sultiame).^{26, 34, 36, 45, 48, 77, 78, 93, 108, 110} Drug–drug interactions were not discussed in the included articles.

ASM discontinuation rates varied across the included studies from less than 5% up to 50%. There were too few studies of each treatment to evaluate whether particular treatments had higher discontinuation rates than others.^{39, 45, 53, 73, 85, 95-98, 108, 114} Reasons given for discontinuation were aggravated seizure severity or frequency,^{95, 97, 108, 114} motor problems,⁵³ vomiting,⁹⁶ and death,^{98, 108} as well as reasons not related to safety, including lack or loss of efficacy^{58, 85, 98, 108} and inability to access the drug.⁹⁸ No comparisons were made across treatments on rates or reasons for discontinuation given the limited data for each treatment.

Non-ASM treatments for LGS included ketogenic diet, corpus callosotomy, VNS, focal resection, deep brain stimulation, antipsychotics, and selective serotonin reuptake inhibitors.^{18, 36, 38, 39, 50, 58, 77, 82, 85, 91, 95, 104, 105, 108-110} Additional treatment modalities are commonly used concurrently with ASMs, with one study from South Korea (n = 68) reporting that 61.8% of individuals were receiving non-ASM treatments concurrently with ASMs.⁸²

Adverse events (AEs) were associated with treatment with ASMs, corpus callosotomy, dietary changes, and VNS. Commonly reported AEs for ASM treatment were increased seizure frequency, behavioral problems, somnolence, nausea or vomiting, anorexia or decreased appetite, weight loss, rash, and dizziness.^{45, 53, 58, 95, 98, 108-110, 114-116}

For VNS, AEs tended to occur soon after surgical implantation of the device and to resolve over time.^{85, 104} Common AEs with VNS included problems related to the respiratory system and speaking, including voice alteration or hoarseness, drooling, and coughing.^{28, 37, 73, 85, 111} For corpus callosotomy, two studies reported AEs, which related to brain connections (e.g., acute disconnection syndrome, aphasia, ataxia, and paresis).^{28, 111} AEs related to dietary therapy were reported in two studies, and were mainly gastrointestinal, including vomiting, diarrhea, constipation, and weight loss, as well as metabolic acidosis.^{56, 96}