Seizures in Pediatric Moyamoya: Risk Factors and Functional Outcomes

Abstract

Introduction:

Children with moyamoya arteriopathy are at high risk for stroke and seizures. Risk factors for seizures and the impact of seizures on neurologic outcomes in children with moyamoya are unknown.

Methods:

This is a single-center retrospective cohort study of children with moyamoya evaluated between 2003 and 2021. Functional outcome was assessed using the Pediatric Stroke Outcome Measure (PSOM). Associations between clinical variables and seizure occurrence were assessed using univariate and multivariable logistic regression. Associations between clinical variables and final PSOM score were assessed using ordinal logistic regression.

Results:

Eighty-four patients met inclusion criteria and 34 (40%) children experienced seizure. Factors associated with seizures included moyamoya disease (versus syndrome; OR 3.43, p=0.008) and presence of infarcts on baseline neuroimaging (OR 5.80, p=0.002). Factors associated with decreased likelihood of experiencing seizures included older age at initial presentation (OR 0.82, p=0.002) and asymptomatic (radiographic) presentation (OR 0.05, p=0.006). Both older age at presentation (AOR 0.80, p=0.004) and incidental radiographic presentation (AOR 0.06, p=0.022) remained significant after adjusting for potential confounders. Seizures were associated with worse functional outcomes as assessed by the PSOM (regression coef. 2.03, p<0.001). This association remained significant after adjusting for several potential confounders (adj. regression coef. 1.54, p=0.025).

Conclusions:

Younger age and symptomatic presentation are associated with increased likelihood of experiencing seizures among children with moyamoya. Seizures are associated with worse functional outcomes. Prospective studies should clarify the role of seizures in moyamoya outcome and how effective seizure treatment modifies this relationship.

INTRODUCTION

Moyamoya is a cerebral arteriopathy characterized by progressive stenosis and occlusion of the distal internal carotid arteries with compensatory formation of lenticulostriate collateral vasculature [1]. Notably, moyamoya is a significant risk factor for ischemic stroke among children, and affected children are at high risk for recurrent strokes over time [2–4]. Moyamoya may occur in association with a systemic condition (such as sickle cell disease, neurofibromatosis-type 1, trisomy 21, history of cranial radiation therapy), termed moyamoya syndrome, or it can occur without a known associated condition, termed moyamoya disease [5]. A plurality of children with moyamoya present with ischemic stroke or transient ischemic attack [6,7]. While seizure is a common manifestation of acute ischemic stroke in children [2,8,9], seizure is also a common manifestation of pediatric moyamoya both within and outside of the setting of acute ischemia [6,7]. The reported frequency of seizure as the presenting symptom of moyamoya in children varies widely among studies, ranging between 2% and 64% [10–12]. Seizures at presentation are most often focal, but roughly 25% of children may present with generalized seizure [12–14]. Children with moyamoya may also experience seizures and epilepsy later in their disease course, possibly due to a remote history of ischemic injury [13].

Despite seizures being a known complication of pediatric moyamoya, there is relatively sparse literature examining risk factors for and long-term prognosis of seizures in this population. Limited observational data from a mixed pediatric and adult East Asian moyamoya disease cohort revealed that patients who had moyamoya onset prior to 3-years-old, greater infarct burden, and cortical infarction were more likely to go on to develop epilepsy. Degree of arterial stenosis as assessed on magnetic resonance angiography (MRA) was not associated with increased risk of epilepsy [15]. However, this cohort had a preponderance of adult patients and did not include patients with moyamoya syndrome [15]. Similarly, a recent North American pediatric cohort inclusive of patients with both moyamoya disease and syndrome revealed that children ≤3 years old at presentation were more likely to present with seizures [7]. However, no study has examined a variety of factors associated with seizures in a pediatric cohort inclusive of moyamoya disease and syndrome. Furthermore, the potential association between seizure and functional outcomes in this population warrants exploration. In this retrospective cohort study of pediatric patients with moyamoya, we assess which clinical and radiographic features are associated with seizures and evaluate for differences in functional outcome among patients with and without seizures.

MATERIALS AND METHODS

This retrospective cohort study was approved by the Johns Hopkins Medicine Institutional Review Board (IRB00196325). Informed consent was waived. Patients with moyamoya arteriopathy seen at our institution between 2003 and 2021 were identified using International Classification of Disease codes, as previously described [6,16]. Patients who received an imaging-confirmed diagnosis of moyamoya arteriopathy between ages 28 days and 18 years were included in this study. Other characteristics of the patients in this cohort have been described in prior studies [6,16].

Medical records were reviewed with attention to demographic factors, comorbid conditions, clinical signs and symptoms, radiographic characteristics, and functional outcomes. Review of records continued into young adulthood if these data were available. Age at diagnosis was calculated using date of birth and either the date of clinical manifestation of moyamoya (if symptomatic) or date of incidental radiographic diagnosis (if asymptomatic). Clinical stroke was defined as an acute-onset focal neurologic deficit lasting longer than 24 hours with associated acute infarction on neuroimaging. Transient ischemic attack was defined as a transient (lasting less than 24 hours) neurologic deficit with no detectable acute infarction on neuroimaging. Seizure was defined as a neurologic event that the treating clinician documented as consistent with seizure.

Radiology records for initial brain magnetic resonance imaging (MRI) and magnetic resonance angiography (MRA) were reviewed. Patients were determined to have infarct(s) at baseline if any infarct, acute or chronic (including silent cerebral infarct), was noted on baseline MRI. Arteriopathy was classified as unilateral or bilateral based on the laterality of signal loss consistent with arterial stenosis on baseline MRA. Patients were determined to have posterior circulation involvement if there was signal loss consistent with stenosis of any posterior circulation artery on baseline MRA. However, to avoid artifactual overestimation of stenosis, patients were only classified as having posterior circulation involvement if this stenosis was confirmed on subsequent MRA or via digital subtraction angiography.

Functional outcomes were assessed using the Pediatric Stroke Outcome Measure (PSOM) [17] as documented at the most recent follow-up visit with a neurologist. If no PSOM score was documented in the medical record, one was assigned retrospectively by a pediatric stroke neurologist (LRS), as has been previously validated [18]. PSOM scores were stratified into four functional outcome categories (normal, mild, moderate, and severe) as previously described [19]. Briefly, this functional outcome category is dependent both on the number of domains affected and the severity of deficits in each domain.

Epilepsy outcomes were assessed using the modified Engel score, as previously described [20,21]. Patients were classified as Class 0 (seizure-free for ≥6 months off anti-seizure medication), Class 1 (seizure-free for <6 months off anti-seizure medication or seizure-free on anti-seizure medication), Class 2 (<1 seizure per month), Class 3 (1–4 seizures per month), Class 4 (5–30 seizures per month, or Class 5 (>30 seizures per month). The clinical note from the most recent encounter with a neurologist at our institution was used to determine the modified Engel classification.

Medical records were reviewed through the most recent encounter at our institution. The time elapsed from the date of initial moyamoya presentation to the date of this encounter was measured. All patients had at least three months of total follow-up time, and all are included in this analysis.

Statistical Methodology

Statistical analysis was performed using Stata 17.0 (StataCorp, College Station, TX). Categorical variables were reported as raw number of patients and percentages. Descriptive statistics for continuous variables were reported using median and interquartile range (IQR). Associations between clinical and imaging factors and experiencing seizure were assessed using univariate logistic regression. These associations were further tested using multivariable logistic regression, and this model included all independent variables with p<0.10 in univariate analysis. Functional outcome categories were graphically depicted using a horizontal bar chart. Distributions of functional outcome categories between children with versus without seizures were compared using ordinal logistic regression. Because there are several clinical and radiographic factors that could confound the relationship between seizures and functional outcome, univariate ordinal logistic regression was used to identify potential confounding factors. Variables with p<0.10 in univariate ordinal logistic regression analysis were used in a multivariable ordinal logistic regression model to assess for a potential independent association between seizures and functional outcome category. Statistical significance was defined as a p-value less than 0.05.

RESULTS

Eighty-four patients with imaging-confirmed moyamoya arteriopathy met inclusion criteria, including 37 (44%) with idiopathic moyamoya disease, 23 (27%) with sickle cell disease, 13 (15%) with neurofibromatosis-type 1, 6 (7%) with trisomy 21, and 5 (6%) with other syndromic associations (Table 1). Thirty-four (40%) children experienced seizures over a median follow-up time of 6.9 years (IQR 3.0–10.9) after initial moyamoya presentation. Of the 34 patients who experienced seizure, the first seizure pre-dated moyamoya diagnosis in five patients (15%), occurring at a median of 8.1 (range 0.4–13.6) years prior to diagnosis. Of these five cases, four first seizures were related to a condition associated with moyamoya (such as neurofibromatosis type 1) and one was a typical febrile seizure felt by the treating clinician to be unrelated to moyamoya. Twenty-three children (68%) experienced their first seizure at the time of moyamoya presentation, including 16 with seizure in the setting of acute stroke; six such patients later developed remote symptomatic seizures. Six children (18%) had no seizures in the acute period but later developed remote symptomatic seizures at a median of 6.7 (range 4.1–96.2) months after initial moyamoya presentation. In total, 19 children (23%) developed remote symptomatic seizures. There was no significant difference in clinical follow-up time between patients with versus without seizure (p=0.58).

Table 1:

Patient characteristics (n=84)

Demographic characteristics
Sex, n (%)
Male	36 (43)
Female	48 (57)
Median years from moyamoya diagnosis to most recent PSOM score (IQR)	4.9 (1.4–9.0)
Median years from moyamoya diagnosis to final visit at institution (IQR)	6.9 (3.0–10.9)
Clinical Characteristics
Associated conditions with moyamoya, n (%)
Idiopathic	37 (44)
Sickle cell	23 (27)
Neurofibromatosis-type 1	13 (15)
Trisomy 21	6 (7)
Cranial radiation/Brain tumor	3 (4)
Hemophilia A	1 (1)
PHACES	1 (1)
Median age at diagnosis in years (range)	6.3 (0.1–17.4)
Initial signs/symptoms at diagnosis, n (%)
Stroke	35 (42)
TIA	12 (14)
Seizure	26 (31)
Radiographic	19 (23)
Other	11 (13)
Presence of acute infarct at initial moyamoya presentation, n (%)	35 (42)
Presence of any infarct at initial moyamoya presentation, n (%)	54 (64)
Laterality of arteriopathy, n (%)
Unilateral	27 (32)
Bilateral	57 (68)
Posterior circulation involvement, n (%)	14 (17)
PSOM functional category, n (%)
Normal	54 (68)
Mild	13 (16)
Moderate	10 (13)
Severe	3 (4)

Abbreviations: PSOM, Pediatric Stroke Outcome Measure

Among the 34 children who developed one or more seizures, 29 (85%) had imaging evidence of acute or chronic infarct at the time of their first seizure. However, five children had no evidence of stroke at the time of their first seizure. Three of these children had identified seizure etiologies unrelated to moyamoya, including a glioma, genetic brain structural abnormalities, and simple febrile seizure. However, two patients had focal seizures with focal or multifocal epileptiform activity on EEG in the setting of a normal brain MRI.

In univariate analysis, there were some notable differences in clinical characteristics between patients with versus without seizures (Table 2). Older age at initial presentation of moyamoya was associated with decreased likelihood of experiencing seizure (OR 0.82, 95% CI 0.7–0.9, p=0.002). Children with moyamoya disease were more likely to experience seizures compared to children with moyamoya syndrome (OR 3.43, 95% CI 1.4–8.5, p=0.008). Moyamoya syndrome associated with neurofibromatosis-type 1 trended toward decreased likelihood of seizure, but this was not statistically significant (OR 0.22, 95% CI 0.0–1.1, p=0.06). While there was no significant association between presenting with overt clinical ischemic stroke and experiencing seizures (OR 2.18, 95% CI 0.9–5.3, p=0.09), children who had evidence of any infarction on baseline MRI (acute or chronic) were more likely to experience seizures (OR 5.80, 95% CI 1.9 – 17.4, p=0.002). There were no significant associations between experiencing seizure and either bilateral arteriopathy or posterior circulation involvement on initial MRA.

Table 2:

Characteristics of patients with versus without seizure

	No Seizure (n=50)	Seizure (n=34)	Odds Ratio	95% CI	p-value
Female Sex	26 (52)	22 (65)	1.69	0.7–4.1	0.25
Idiopathic MMD	16 (32)	21 (62)	3.43	1.4–8.5	0.008
Sickle cell	17 (34)	6 (18)	0.42	0.1–1.2	0.10
Neurofibromatosis-type 1	11 (22)	2 (6)	0.22	0.0–1.1	0.06
Trisomy 21	3 (6)	3 (9)	1.52	0.3–8.0	0.62
Cranial radiation	2 (4)	1 (3)	0.73	0.1–8.4	0.80
Median (range) age at diagnosis, years	7.49 (0.1–17.4)	4.79 (0.5–14.5)	0.82	0.7–0.9	0.002
Initial signs/symptoms at diagnosis
Stroke	17 (34)	18 (53)	2.18	0.9–5.3	0.09
TIA	8 (16)	4 (12)	0.70	0.2–2.5	0.59
Radiographic	18 (36)	1 (3)	0.05	0.0–0.4	0.006
Other	7 (14)	4 (12)	0.82	0.2–3.0	0.77
Presence of any infarct	25 (50)	29 (85)	5.80	1.9–17.4	0.002
Bilateral arteriopathy	33 (66)	24 (71)	1.23	0.5–3.2	0.66
Posterior circulation involvement	8 (16)	6 (18)	1.13	0.4–3.6	0.84

All values are n (%) unless otherwise noted.

Abbreviations: MMD, moyamoya disease

To adjust for potentially confounding factors, multivariable analysis including all independent variables with a p-value <0.1 in univariate analysis was performed (Table 3). Age at moyamoya presentation remained significantly associated with seizures, with younger children being more likely to experience seizures (AOR 0.80, 95% CI 0.7–0.9, p=0.004). Children with an incidental radiographic (as opposed to symptomatic) presentation of moyamoya were less likely to experience seizures (AOR 0.06, 95% CI 0.0–0.7, p=0.02). There was a trend toward presence of infarcts on baseline MRI being associated with increased likelihood of seizure, but this did not reach statistical significance (AOR 4.34, 95% CI 0.4–24.2, p=0.08). Moyamoya disease and comorbid neurofibromatosis-type 1 were not associated with likelihood of experiencing seizures after adjusting for potentially confounding factors.

Table 3:

Multivariable analysis of patients with versus without seizure

	Adjusted Odds Ratio†	95% CI	p-value
Idiopathic MMD	1.13	0.3–4.1	0.85
Neurofibromatosis-type 1	0.25	0.0–2.5	0.24
Age at presentation	0.80	0.7–0.9	0.004
Acute stroke at presentation	0.27	0.1–1.3	0.11
Radiographic presentation	0.06	0.0–0.7	0.02
Presence of infarcts on baseline imaging	4.34	0.4–24.2	0.08

Abbreviations: MMD, moyamoya disease

^†Multivariable model includes all clinical variables listed in the first column of this table

Seizures were associated with less favorable functional outcomes, as assessed by PSOM scores at most recent follow-up visit (Figure 1). PSOM scores were available for 80 (95%) children. PSOM scores were stratified into different functional outcome categories (normal, mild, moderate, severe). Overall, 13 (16%) children had a moderate/severe functional outcome, as assessed by the PSOM. Of these 13 children, 11 (32% of the seizure group) experienced one or more seizures, while only two did not experience one or more seizures. Children with seizures had worse outcomes as assessed by the PSOM compared to children without seizures (regression coefficient 2.03, 95% CI 1.0 – 3.1, p<0.001). Even after adjusting for comorbid trisomy 21, age at moyamoya presentation, acute stroke at presentation, incidental radiographic (asymptomatic) presentation, presence of infarcts on baseline MRI, presence of bilateral arteriopathy at time of presentation, and posterior circulation involvement at time of presentation (Table S1), seizure remained independently associated with less favorable outcome (regression coefficient 1.54, 95% CI 0.2 – 2.9, p=0.03).

Figure 1, Pediatric Stroke Outcome Measure by Seizure Occurrence:

Horizontal bar chart depicting functional outcome categories between patients without (top) versus with (bottom) seizures. When compared via ordinal logistic regression, seizures are associated with overall less favorable functional outcome (regression coefficient 2.03, p<0.001). This remains significant after adjusting for potential confounding variables (adjusted regression coefficient 1.54, p=0.025). Functional outcome was assessed using Pediatric Stroke Outcome Measure (PSOM), and categories were determined using the criteria described by Felling et al. [19].

Most children’s seizures were well-controlled by medical therapy, though some children developed severe epilepsy (Table 4). Among the 34 patients who had one or more seizures, 32 (94%) were started on anti-seizure medication. Among the 32 children started on anti-seizure medication, the most common first line agent was levetiracetam (14 children, 44%). Of the 34 children who had one or more seizures, the vast majority achieved seizure freedom (modified Engel Class 0 or 1, 30 children, 88%). Two children were classified in modified Engel Class 5 (>30 seizures per month).

Table 4:

Seizure Control (n=34)

	n (%)
Started on ASM	32 (94)
First-line ASM (n=32)
levetiracetam	14 (44)
carbamazepine	5 (16)
phenobarbital	5 (16)
oxcarbazepine	3 (9)
othera	5 (16)
Second-line ASM (n=18)a
oxcarbazepine	7 (39)
carbamazepine	3 (17)
otherb	8 (44)
Maximum number of ASMs required at one time
0	2 (6)
1	22 (65)
2	7 (21)
3 or more	3 (9)
Modified Engel score
0	15 (44)
1	15 (44)
2	0 (0)
3	2 (6)
4	0 (0)
5	2 (6)

Abbreviations: ASM, anti-seizure medication

^aphenytoin (2), lamotrigine (1), topiramate (1), valproic acid (1)

^blacosamide (2), levetiracetam (2), topiramate (2), valproic acid (1), zonisamide (1)

DISCUSSION

In this cohort of children with moyamoya, moyamoya disease (versus syndrome), young age at presentation, symptomatic presentation (versus incidental radiographic), and infarcts on baseline imaging were associated with experiencing seizures. In multivariable analysis, young age at presentation and symptomatic moyamoya presentation were independently associated with experiencing seizures. Children who experienced seizures had overall less favorable functional outcomes, and this association remained significant after adjusting for a host of clinical and radiographic factors. Taken together, these findings suggest that seizures may occur more frequently in children who are younger at presentation and may be associated with a more aggressive disease course overall. Our findings are consistent with previous studies demonstrating an association between young age at presentation and risk of seizures [7,15]. Interestingly, there was marked variability with respect to severity of seizures. While most children’s seizures were well-controlled on or off anti-seizure medications, there were several cases of severe epilepsy in our cohort.

There are several reasons why children with moyamoya may be prone to seizures. The cumulative risk of developing remote symptomatic seizures after arterial ischemic stroke is as high as 33% among children, which is much higher than among adults [22], suggesting that ischemic injury to the developing brain may be more epileptogenic than ischemic injury later in life. In addition, children with moyamoya are subject to chronic cerebral hypoperfusion, which may predispose affected children to developing seizures in the absence of or in concert with cerebral infarction.

There are also many possibilities for the less favorable functional outcomes noted among children who experienced seizure. Independent of moyamoya, post-stroke epilepsy in children has been associated with worse functional outcome, cognitive impairment, and worse parent-reported health status[23–25]. In moyamoya, the potential for less favorable outcomes is compounded by both chronic cerebral hypoperfusion and risk for recurrent strokes over time. It is possible that more severe moyamoya leads to more significant infarct burden, causing higher risk of both seizures and neurologic impairment. Repeated seizures over time and/or prolonged seizures may also alter neural networks, causing cognitive dysfunction and/or hampering stroke recovery. Seizures also increase metabolic demand and could cause ischemic injury in a patient with severe moyamoya whose vasculature is unable to compensate with increased blood flow. It is also possible that medications used to control seizures, especially sedating and/or older generation medications used in this cohort, may contribute to less favorable cognitive outcomes.

There are important limitations of this study. Its retrospective nature makes interpretation of episodic neurologic events challenging. In many cases, it is unclear whether a transient neurologic event represents a seizure, a transient ischemic attack, or a movement disorder. In a retrospective study, we lack the ability to use objective criteria to adjudicate these events. To minimize bias, all events were adjudicated as documented by the treating physician. Finally, though robust for a single-center cohort study of a rare disease, the small sample size limits the reliability of multivariable analyses when adjusting for multiple factors.

The results of this study lead to several questions for future investigation. Future studies should analyze the effects of epilepsy severity and seizure control on clinical outcomes. Whether recurrent seizures increase the risk of cerebral ischemia due to mismatch between blood flow supply and increased metabolic demand during a seizure is an important area for future investigation. As abnormal EEG characteristics in children can improve following surgical revascularization [13], future prospective studies evaluating the role of revascularization surgery for seizure control are warranted. Furthermore, prospective studies investigating the relationship between infarct size and location and associated risk of epilepsy are warranted. Advanced neuroimaging connectivity analyses in children with moyamoya with and without seizures may also help understand the relationship between seizures, ischemic injury, and disruption of neural networks. Finally, studies on the relationship between seizures and poor functional outcomes could help disentangle whether seizures are directly responsible for worse outcomes or whether they are merely a marker for more severe arteriopathy.

In conclusion, young age at presentation and symptomatic presentation of moyamoya are independently associated seizures. Seizures are independently associated with less favorable functional outcomes, as assessed by the PSOM. Further study on this important topic will help inform management of these complex children and will aid in counseling families about their child’s long-term prognosis.

Data Availability:

Data may be made available upon reasonable request and in accordance with institutional data sharing policies.