Risk Factors for White Matter Lesion Burden in Adults with Focal Epilepsy

Introduction

Focal epilepsy has traditionally been considered a disorder originating from a localized epileptogenic focus, but growing evidence now suggests that it is fundamentally a large-scale brain network disorder. ¹ White matter tracts serve as the anatomical substrate for transmitting neural signals in the brain; thus, their structural integrity and functional status are essential for maintaining normal brain function. Neuroimaging studies consistently reveal widespread white matter changes in epilepsy patients, manifesting at the microstructural level as increased diffusivity and reduced fractional anisotropy (FA) and at the macroscopic network level as decreased global and local efficiency.^2,3 Notably, these changes are not limited to the epileptogenic focus; they are also observed in white and gray matter regions distant from it, reflecting the widespread nature of the epileptic network. ⁴ Animal models further confirm that recurrent focal seizures can cause widespread structural damage to major white matter tracts in both hemispheres. ⁵

Among these white matter changes, white matter hyperintensities (WMH) are among the most visible and easily quantifiable markers on conventional magnetic resonance imaging (MRI). The prevalence of WMH is higher in patients with focal epilepsy than in healthy individuals, with a complex and varied pathological basis. In specific etiologies, such as focal cortical dysplasia (FCD), WMH is directly associated with developmental pathological changes, including poor myelination and axonal fiber loss.^4,6 For most patients, the reduction in FA, which indicates myelin loss or axonal injury, may result from secondary Wallerian degeneration triggered by recurrent seizures.^7,8 Furthermore, Antiseizure Medications(ASMs) may contribute to this process. Studies have shown that patients treated with enzyme-inducing ASMs carry a significantly higher burden of WMH. ⁹ These findings suggest that both the epilepsy itself and its treatment contribute to the formation and progression of WMH, with epilepsy-related white matter microstructural damage potentially developing into visible WMH over time. ¹⁰

There is a bidirectional interaction between white matter networks and epilepsy. On the one hand, white matter serves as a key pathway for the conduction of epileptic discharges, and changes in its diffusion parameters can assist in localizing the epileptogenic focus, especially in MRI-negative patients. ¹¹ Conversely, lesions in white matter can serve as an initiating factor for epilepsy. WMH, a marker of cerebral small vessel disease, may increase the risk of late-onset epilepsy by affecting cortical microstructure or causing microinfarcts. ¹² This bidirectional relationship underscores that WMH is not only a consequence of epilepsy but also a potential causal and modifying factor, with its burden likely closely tied to disease severity and clinical outcomes. ²

Although the importance of WMH in epilepsy is increasingly recognized, current research primarily addresses descriptive phenomena. There remains a lack of systematic, multivariable analyses identifying which factors—such as epilepsy duration, seizure frequency, seizure duration, antiseizure medication types, and vascular risk factors—are independent risk factors for increased WMH burden and of assessing their relative importance. This knowledge gap limits our ability to identify high-risk patients and develop targeted management strategies.

Accordingly, this cross-sectional study aims to systematically investigate the clinical risk factors associated with WMH burden in patients with focal epilepsy. The findings are expected to elucidate key drivers of epilepsy-related white matter injury, to provide a theoretical foundation for future research on the adjunctive localization of epileptogenic foci and on interventions aimed at mitigating white matter damage, and to offer insights into the involvement and disruption of white matter tracts within the epileptic brain network. ⁵

Materials and Methods

Research Participants

This retrospective analysis included 60 adults with focal epilepsy who met the inclusion and exclusion criteria between October 2024 and October 2025. Patients aged 14 to 18 years were included because their brain development was considered to have largely reached maturity. The screening flowchart is shown in Figure 1.

OPEN IN VIEWER

Figure 1.

Flow diagram of the enrollment process.

Inclusion criteria were as follows: (1) age 14–70 years, with a diagnosis of focal epilepsy confirmed by a multidisciplinary team according to the 2017 International League Against Epilepsy guidelines; (2) completion of a 3 T epilepsy-protocol MRI and 17-h video electroencephalography (EEG) monitoring that included at least one full sleep cycle; and (3) availability of detailed seizure semiology, including seizure videos provided by patients or caregivers, seizure frequency before the visit, and complete long-term video EEG and 3 T MRI data.

Exclusion criteria were as follows: (1) autoimmune diseases, chronic inflammatory disorders, or other chronic progressive systemic conditions; (2) cerebrovascular and related disorders, including migraine, ischemic stroke, hypertension, intracerebral hemorrhage, and arteriovenous malformations; (3) systemic malignancies or intracranial tumors; and (4) a history of severe traumatic brain injury or previous neurosurgical procedures; (5) multifocal epilepsy.

Results

Risk Factors Associated with WMH Burden

Univariable ordinal logistic regression analysis revealed that age (p = 0.006), seizure duration (p = 0.024), seizure frequency of at least once per day (p = 0.011), and interictal epileptiform discharges (IEDs) (p < 0.001) were significantly associated with WMH severity. Sex, disease duration, seizure frequency of ≤1 per week, symptomatology, presence of MRI abnormalities, and antiseizure medication use showed no significant associations (all p > 0.05). Variables with p < 0.05 in univariable analysis were entered into a multivariable ordinal logistic regression model. Multivariable analysis identified age (OR = 1.06, 95% CI: 1.01-1.10, p = 0.012), seizure duration (OR = 1.60, 95% CI: 1.13-2.24, p = 0.007), and the presence of IEDs (OR = 5.83, 95% CI: 1.71-19.87, p = 0.005) as independent predictors of WMH severity. Detailed results of the univariable and multivariable analyses are presented in Table 3.

OPEN IN VIEWER

Table 1.

Demographic and Clinical Characteristics.

Variables	n (%), Mean ± SD, Max/Min/Mdn
Sex	F/M: 32/28
Age	36.55 ± 15.96 (Years)
Age at onset	26.87 ± 14.677 (Years)
Duration of disease	Min: 4, Max: 600, Mdn: 54 (Months)
Duration of episode	2.69 ± 1.81（Mins）
MRI Abnormalities
Hippocampal Atrophy or Sclerosis	22 (36.67%)
Small Encephalomalacic Foci	3 (5.00%)
Arachnoid Cyst	2 (3.33%)
Focal cortical dysplasia (FCD)	2 (3.33%)
Schizencephaly	1 (1.67%)
Non-lesional MRI	30 (50.00%)
Seizure types
Clonic	2 (3.33%)
(ImpairedAwareness)Automatisms	6 (10%)
GTCS	51 (85.00%)
Sensory	1 (1.67%)
ASM use
monotherapy	34 (56.67%)
polytherapy	17 (28.33%)
None	9 (15.00%)

OPEN IN VIEWER

Table 2.

Participants’ Demographic Characteristics and the Variables are Categorized as Follows.

Characteristics	Score	Grouping	Count n = 60	Ratio %
Gender	0	Female	32	53.3%
	1	Male	28	46.7%
Course of disease	0	≤3months	10	16.7%
	1	>3 to ≤12 months	8	13.3%
	2	>12 months	42	70%
Smoking history	0	Never	59	98.3%
	1	Current	1	1.7%
Drinking history	0	Never	60	100%
	1	Current	0	0%
Seizure frequency	0	<1 times per week	21	35%
	1	>1 and ≤7 times per week	32	53%
	2	≥1 times per day	7	11.7%
Semiology	0	Non-GTCS	8	13.3%
	1	GTCS	52	86.7%
Oral medication use	0	Non-enzyme-inducing ASMs	17	28.3%
	1	Enzyme-inducing ASMs	14	23.3%
	2	No ASMs or Herbal Medicine	29	48.3%
MRI Finding	0	Normal	30	50%
	1	Abnormal	30	50%
IEDs	0	Absent	34	56.7%
	1	Present	26	43.3%
Modified Scheltens Score	0	Score≤1	17	28.3%
	1	1 < Score <5	24	40%
	2	5≤ Score < 10	8	13.3%
	3	Score≥10	11	18.3%

GTCS:Generalized Tonic-Clonic Seizures, IEDs:Interictal Epileptiform Discharge

OPEN IN VIEWER

Table 3.

Factors Associated with White Matter Hyperintensities in Univariate and Multivariate Analyses.

Variables	Univariate Analysis					Multivariate Analysis
	β	S.E	t	P	OR (95%CI)	β	S.E	t	P	OR (95%CI)
Gender
0
1	0.22	0.47	0.46	0.643	1.24 (0.49 ∼ 3.14)	0.93	0.59	1.58	0.115	2.53 (0.80 ∼ 7.99)
Course of disease
0
1	−0.17	0.88	−0.20	0.844	0.84 (0.15 ∼ 4.76)	−0.87	1.09	−0.79	0.427	0.42 (0.05 ∼ 3.58)
2	0.19	0.67	0.28	0.776	1.21 (0.32 ∼ 4.53)	−0.23	0.89	−0.26	0.797	0.80 (0.14 ∼ 4.52)
Seizure frequency
0
1	−0.84	0.53	−1.59	0.112	0.43 (0.15 ∼ 1.22)	0.20	0.71	0.28	0.781	1.22 (0.30 ∼ 4.88)
2	−2.19	0.86	−2.55	0.011	0.11 (0.02 ∼ 0.60)	−0.65	1.03	−0.63	0.529	0.52 (0.07 ∼ 3.95)
Semiology
0
1	−0.34	0.67	−0.50	0.616	0.71 (0.19 ∼ 2.66)	−1.05	0.94	−1.13	0.259	0.35 (0.06 ∼ 2.18)
Oral medication use
0
1	0.18	0.65	0.27	0.786	1.19 (0.34 ∼ 4.22)	0.35	0.74	0.47	0.638	1.41 (0.33 ∼ 5.97)
3	−0.17	0.55	−0.32	0.752	0.84 (0.29 ∼ 2.45)	0.08	0.70	0.12	0.908	1.08 (0.27 ∼ 4.28)
MRI Finding
0
1	−0.01	0.47	−0.03	0.975	0.99 (0.39 ∼ 2.48)	−0.37	0.63	−0.59	0.558	0.69 (0.20 ∼ 2.38)
IEDs
0
1	1.76	0.53	3.31	<.001	5.84 (2.05 ∼ 16.60)	1.76	0.63	2.82	0.005	5.83 (1.71 ∼ 19.87)
Age	0.05	0.02	2.75	0.006	1.05 (1.01 ∼ 1.09)	0.06	0.02	2.51	0.012	1.06 (1.01 ∼ 1.10)
Duration of episode	0.30	0.13	2.25	0.024	1.35 (1.04 ∼ 1.75)	0.47	0.17	2.69	0.007	1.60 (1.13 ∼ 2.24)

OR: Odds Ratio, CI: Confidence Interval

The forest plot from the multivariable logistic regression analysis is presented in Figure 2. Age, seizure duration, IEDs, and seizure frequency of at least once per day are independent risk factors for WMH burden. After adjusting for age, the odds ratios for seizure duration and IEDs remained greater than 1, indicating an increased risk of WMH burden.

OPEN IN VIEWER

Figure 2.

Forest plot of risk associations for white matter hyperintensities.

Regional Differences in WMH Scores

Regional differences in WMH burden were assessed using the modified Scheltens scale. Inter-rater reliability, assessed using Cohen's kappa with quadratic weights, was substantial (κ = 0.82, 95% CI: 0.74-0.89). As WMH scores did not follow a normal distribution (Shapiro–Wilk test, p < 0.05), regional differences in WMH burden were compared using the Kruskal–Wallis test. Post hoc pairwise comparisons with the Bonferroni correction revealed that frontal lobe WMH burden was significantly higher than that in the temporal lobe (adjusted p = 0.004) and the occipital lobe (adjusted p = 0.022). No other significant differences were observed among regions (adjusted p > 0.05). Detailed pairwise comparison results are presented in Table 4. The absolute counts and percentages of WMH lesions for each brain region are presented in Table 5.

OPEN IN VIEWER

Table 4.

Regional Variation in White Matter Hyperintensity Burden.

Sample 1-Sample 2	Adjusted P Valuea
Temporal-Occipital	1.000
Temporal-Parietal	1.000
Temporal-Periventricular	.898
Temporal-Frontal	.004
Occipital-Parietal	1.000
Occipital-Periventricular	1.000
Occipital-Frontal	.022
Parietal-Periventricular	1.000
Parietal-Frontal	.113
Periventricular-Frontal	.676

OPEN IN VIEWER

Table 5.

The Absolute Counts and Percentages of WMH Lesions for Each Brain Region.

WMH
Location	juxtacortical	72
	subcortical U-fibers	168
	Deep non-limbic cortical regions	142
	subcortical limbic regions	117
	periventricular	43
	corpus callosum	8
lobar WMH
>3mm	231	42%
≤3mm	319	58%
	total	550

Case Presentation

MRI images from two patients with WMH, together with their clinical information, are shown in Figure 3.

OPEN IN VIEWER

Figure 3.

A: 35-year-old male, left hippocampal sclerosis, FBTCS, frequency once/week. B: 50-year-old male, right hippocampal atrophy, FBTCS, frequency twice/week.

Discussion

Research on WMH has predominantly focused on dementia, with relatively few studies conducted in the context of epilepsy. WMH are closely associated with disruption of white matter structural integrity and reduced interhemispheric functional connectivity, with the latter linked to declines in processing speed and executive function. ¹³ Aging is an independent factor influencing both the prevalence and severity of WMH, with age-related risk factors contributing to its development.^14,15 A greater baseline WMH volume increases the risk of cognitive impairment and dementia. ¹⁶ WMH volume has also been associated with essential tremor. ¹⁷ The location and severity of WMH affect cognitive deficits and may provide clues to the underlying etiology; frontal and parietal WMH have been linked to impairments in executive function and memory.^18–21

Evidence suggests that patients with epilepsy are more likely to exhibit WMH, particularly those treated with enzyme-inducing antiseizure medications. ⁹ White matter tract damage has been observed in patients with temporal lobe epilepsy and hippocampal sclerosis. ¹¹ In children and adolescents with temporal lobe epilepsy, bilateral white matter involvement may reflect developmental brain plasticity and the widespread nature of epileptic networks. ³ The extent of WMH in early to middle adulthood may predict the subsequent development of late-onset epilepsy ¹² and is also associated with antiseizure medication use. ⁹

In the present study, no association was found between enzyme-inducing antiseizure medications and WMH. The frontal lobe was the most common site of WMH in our cohort; whether this spatial distribution relates to seizure propagation remains uncertain and warrants further investigation. Age, IEDs, and seizure duration were identified as independent risk factors for WMH. Based on current evidence, a spatial correlation between IEDs and WMH cannot be established, primarily because the sites of white matter pathology do not correspond predictably to IED localization and because IEDs are limited by poor spatial resolution, making it difficult to accurately reflect specific anatomical locations.

Further investigation of spatial associations will require more advanced technical approaches, representing a direction for our ongoing research. Notably, the presence of generalized tonic-clonic seizures was not identified as a risk factor for WMH in this study—a finding that diverges from conventional understanding and warrants validation in larger cohorts. In the present study, 42% of WMH were >3 mm in diameter, whereas 58% were round or oval lesions measuring <3 mm. The lesions were predominantly located in subcortical U-fibers, deep non-limbic cortical regions, and deep limbic cortical regions, whereas periventricular and juxtacortical areas were also relatively frequently involved, and the corpus callosum was rarely affected. This distribution pattern differs markedly from that of WMH attributed to other etiologies, suggesting an underlying mechanism distinct from that of cerebrovascular disease. This may also explain the low detection rate on conventional slice-thickness MRI, as WMH in this cohort were relatively small in volume, posing challenges for automated detection; consequently, manual counting was employed rather than volumetric analysis.

Our findings support the view that clinical investigations should consider not only the size but also the location of WMH. ²² A meta-analysis of studies on temporal lobe epilepsy suggested that WMH are predominantly confined to unilateral subcortical regions, particularly the posterior cerebral cortex or areas surrounding preexisting lesions. ⁴ Recurrent seizures may induce axonal injury in ipsilateral fiber tracts and cause associated secondary Wallerian degeneration. ⁴ In patients with refractory temporal lobe epilepsy, secondary white matter degeneration may occur in the corpus callosum, with pathological changes extending beyond the temporal lobe to involve the contralateral limbic system, thalamus, and remote extratemporal structures.^7,8

WMH observed on MRI can be considered indicative of white matter tract injury. Whether the spatial distribution of these lesions corresponds to critical nodes within the electrical conduction network remains unexplored, and the underlying biological mechanisms of such injury have not been elucidated. The relationship between white matter damage and seizure activity also remains unclear—whether the injury is caused by epileptic electrical activity, ischemia, hypoxia, or immune-mediated processes remains to be determined. Moreover, whether white matter injury can recover following seizure control is unknown. Additionally, no studies have investigated whether the distribution of white matter damage may aid in localizing epileptogenic foci. The present study provides only a preliminary exploration of the association between IEDs and WMH burden; further in-depth investigation is needed to clarify the risk factors for WMH and their relationship with epileptic electrical conduction pathways. Strict inclusion and exclusion criteria were applied in this study to minimize confounding factors and ensure the reliability of the findings. Nevertheless, WMH remained associated with age, and after adjusting for age, a significant correlation between WMH and seizure duration persisted.

This study has several limitations. First, its retrospective design and relatively small sample size may limit statistical power to detect differences between groups. Although some studies have suggested that deep learning methods do not outperform conventional approaches for WMH identification and segmentation, ²³ this was not examined in the present study. The modified Scheltens scale was used, which demonstrates good reliability and validity; it enables assessment of both lesion size and location as well as the number of deep white matter lesions and shows good correlation with WMH volume. This scale may be particularly sensitive in longitudinal observation of WMH progression and its relationship with clinical manifestations.

Future studies may consider using machine learning for automated segmentation and volumetric analysis, combined with fiber tractography to localize affected tracts and map them to standard space. ²⁴ However, some have suggested that spatial normalization may introduce inaccuracies in volumetric measurement. Future work should also incorporate source localization analysis using scalp EEG to enhance study quality and reproducibility. Scalp EEG abnormalities show some degree of spatial deviation from their underlying anatomical structures; increasing electrode density may improve spatial correspondence. Future studies with larger sample sizes could employ spatial overlap analysis using radiomics and source localization techniques to further investigate the mechanisms underlying WMH formation in patients with epilepsy and its relationship to the epileptic network.