Voxel-Based Mapping of Seizure Risk in Meningioma Identifies the Frontoparietal Interface and Tumor Microenvironment as Key Predictors

Abstract

BACKGROUND AND OBJECTIVES:

Brain invasion (BINV) and peritumoral edema (PTBE) have repeatedly been implicated as major contributors to meningioma-related seizures (MRS). By contrast, the influence of the precise tumor location on seizure risk is still insufficiently characterized. In this study, we systematically investigated clinical and imaging-derived predictors of MRS, integrating them with spatial analyses using voxel-based lesion symptom mapping (VLSM).

METHODS:

We reviewed clinical variables—including seizure history, histopathological subtype, World Health Organization (WHO) grade, BINV, and extent of resection—alongside MRI features such as PTBE, tumor dimensions, and anatomic location in a cohort of 532 meningioma patients. Significant predictors of MRS were identified using multivariate stepwise regression. These predictors were subsequently entered into a multivariate VLSM framework to determine their anatomic distribution.

RESULTS:

A total of 93 of 532 patients (17%) experienced seizures prior to surgery. BINV, PTBE, and tumor location emerged as independent predictors of preoperative MRS. VLSM revealed the frontoparietal transition zone (FPTZ) as the region most strongly associated with seizures linked to BINV, whereas PTBE predominantly predicted MRS in frontal convexity and sphenoid wing tumors. Although WHO grade and histological classification displayed distinct spatial patterns, neither was directly related to MRS. Postoperatively, persistent seizure activity occurred in 6% of patients, again implicating the FPTZ as a major risk region. Seizures associated with PTBE were more likely to resolve following tumor removal.

CONCLUSION:

Tumor-intrinsic characteristics such as histology, WHO grade, and lesion size play only a limited role in the development of MRS. Instead, BINV, PTBE, and the anatomic location of the meningioma are the principal determinants of seizure risk. Our findings underscore the FPTZ as a critical region linking tumor invasiveness, seizure generation, and postoperative seizure control.

ABBREVIATIONS:

AED

antiepileptic drugs

BINV

brain invasion

DAZ

dural attachment zone

EMT

epithelial-to-mesenchymal transition

FPTZ

frontoparietal transition zone

MRS

meningioma-related seizures

PostS

postoperative seizures

PTBE

peritumoral brain edema

VLSM

voxel-based lesion symptom mapping

WHO

World Health Organization

WHO2

World Health Organization grade 2.

Meningiomas are the most prevalent primary central nervous system tumors, constituting approximately 36% of all cases.¹ Although these tumors are typically benign and extra-axial, 10% to 50% of patients experience meningioma-related seizures (MRS).^2-7 Tumor removal results in seizure freedom in 50% to 90%.^8,9 However, 30% to 60% of patients do not obtain seizure freedom or develop new-onset seizures postoperatively.^9-13 MRS is a significant cause of morbidity.¹⁴

MRS result from multiple factors that increase the excitability of the peritumoral cortex. Extensive research has identified mass effect, brain invasion (BINV), and peritumoral brain edema (PTBE) as positive predictors of MRS.^2-4,15 However, these factors are not independent. PTBE is well known to correlate with BINV.^9,16 In turn, BINV is an independent grading criterion for WHO2 meningiomas.¹⁷ High-grade meningiomas are associated with larger tumor volume¹⁸ and considered to exert a higher risk of MRS.^6,9,16 However, the extent to which World Health Organization (WHO) grading influences MRS remains uncertain and may depend on the specific subtype.¹⁹

We hypothesized that tumor location is a major and underestimated factor in MRS. First, certain locations (e.g., the frontal convexity) may allow larger tumor volumes to develop before neurological symptoms appear. Second, the region-specific vulnerability of the peritumoral cortex may influence the likelihood of MRS. For example, meningiomas arising from nonskull base locations are more likely to be associated with MRS.^2,7,12,20 Third, tumor location is also related to WHO grading and meningioma subtypes.^17,21 In addition, nonskull base meningiomas are more likely to present higher WHO grading.^18,22-24

The location of intracranial meningiomas is generally described by their dural attachment zone (DAZ). In large tumors, DAZ can be obscured, and in other cases (such as convexity meningiomas), DAZ description can be imprecise. Voxel-based lesion symptom mapping (VLSM) offers a solution to these challenges. This technique geometrically normalizes individual MRI scans to a common space, allowing for an objective inter-individual statistical analysis of tumor locations in relation to the associated symptoms or treatment outcome. VLSM has already been successfully applied to describe glioblastomas, brain metastases and primary central nervous system lymphoma,^25-29 as well as epileptogenic lesions.³⁰ There are only few studies applying VLSM of intracranial meningiomas.^31-34 However, none of them has evaluated the relationship between DAZ and MRS. Thus, this study aims to evaluate the predictors of MRS considering the exact spatial profile.

METHODS

Patient Population

This retrospective observational single-center study enrolled 1175 patients who underwent surgical treatment of intracranial meningioma between 2007 and 2017 at our institution. In total, 541/1175 (46%) cases met the inclusion criteria defined as sufficient quality of preoperative MRI, no previous treatment, and complete clinical information about preoperative and postoperative seizures (PostS). Nine patients with intraventricular meningioma were excluded (Supplemental Digital Content 1, Figure 1, http://links.lww.com/NEU/F225). The clinical data were acquired from medical records including age at diagnosis, gender, histopathologic findings, extent of resection (i.e., Simpson grade) and tumor location. Since this was a retrospective patient cohort, the clinical data were reviewed by blinded experts and, if necessary, reclassified to ensure data comparability. Histopathological diagnosis was reclassified according to the WHO 2021 classification.¹⁷ Seizure phenotype was classified according to the International League Against Epilepsy 2017 classification (i.e., focal or generalized/unknown).³⁵ Seizures that led to the initial diagnosis of the meningioma were classified as MRS. For statistical purposes, MRS were included independent of the seizure phenotype. After surgery, antiepileptic drugs (AED) are continued and tapered off if the patient has been seizure-free for at least 1 year, and electroencephalography shows no abnormalities.

Early perioperative seizures are often acute symptomatic (e.g., edema, hematoma, surgical trauma) and related to transient perioperative factors rather than persistent epileptogenicity. To avoid this bias, only PostS occurring >3 months after surgery were considered for further analysis. It should be noted that this cutoff exceeds the time frame commonly applied for acute symptomatic seizures attributed to metabolic, toxic, ischemic, infectious, or inflammatory etiologies.³⁶ In addition, we excluded seizures occurring after the detection of recurrences or the initiation of second-line treatments (e.g., radiotherapy). PTBE was defined radiologically as the presence of any fluid-attenuated inversion-recovery or T2 hyperintensity around the meningioma regardless of size. BINV was defined based on (1) histopathological reports describing infiltration of brain tissue beyond the arachnoid layer and/or (2) explicit intraoperative description of pial or cortical infiltration. This study was approved by the local ethics committee (No. 702/2024B02) and performed in accordance with the Declaration of Helsinki. Patients provided written informed consent. The results are reported following the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.³⁷

MRI

The MRI was conducted at our institution using 1.5 T magnetic resonance scanner (Siemens Healthineers) including a high-resolution T1-weighted contrast-enhanced Magnetization-Prepared Rapid Gradient-Echo sequence (slice thickness: 1.0 mm; repetition time: 2300 ms; echo time: 3.51 ms; inversion time: 1100; flip angle: 8°; pixel bandwidth: 130; pixel spacing: 1/1 mm; matrix: 256 × 256). However, many segmentations were based on preoperative MRI performed externally, and thus, the parameters varied considerably among patients. The most recent scan obtained before surgery was used for analysis with a median interval between MRI and surgery of 30 days (IQR 11-66).

VLSM

All MRI data sets originally stored in Digital Imaging and Communications in Medicine format were first converted into Neuroimaging Informatics Technology Initiative files using dcm2niix.³⁸ Image preprocessing was performed with Statistical Parametric Mapping software (SPM12, Wellcome Centre for Human Neuroimaging, University College London; https://www.fil.ion.ucl.ac.uk/spm/docs/) in combination with MATLAB (R2024a, MathWorks). Individual MRI scans were coregistered and spatially normalized to the MNI152 standard space (Montreal Neurological Institute, McGill University) (Figure 1). The normalized volumes were then resampled to an isotropic resolution of 1 × 1 × 1 mm. Tumor boundaries were delineated semiautomatically using MATLAB's fast-marching segmentation algorithm (imsegfmm.m). Manual corrections to the tumor mask were made in MRIcron (https://www.nitrc.org/projects/mricron/). All tumor masks were mirrored to the right hemisphere of the brain. DAZ was acquired by intersecting the tumor mask with a template representing the dura, falx, and tentorium. This template was generated based on a publicly available high-resolution computed tomography template³⁹ and falx/tentorium masks provided by Brain Biomechanics Imaging Resources Project (https://www.nitrc.org/projects/bbir).

FIGURE 1.

Spatial distribution of intracranial meningioma. A, Workflow for voxel-based lesion symptom mapping. Patients' individual Neuroimaging Informatics Technology Initiative data were registered to a 1-mm isotropic, high-resolution, T1-weighted MRI template provided by the MNI (the MNI152). The contrast-enhancing lesions were semi-automatically segmented and flipped so that all tumor masks were concentrated onto the left side of the brain. Tumor masks were intersected with a template representing the dura and falx to detect the DAZ of the meningioma. This information was used for the voxel-wise lesion symptom mapping analysis. B, Lesion density (frequency) maps, representing the spatial distribution of meningioma DAZ in the present cohort (n = 534), indicated hotspots at the area of the middle one-third of the superior sagittal sinus, the falx, along frontoparietal transition zone (eg, the coronal suture) and skull base structures such as parasellar, sphenoid wing, and petroclival regions. There were only few lesions observed at the parieto-occipital convexities. DAZ, dural attachment zone; MNI, Montreal Neurological Institute.

VLSM was based on a univariable linear regression model and one-tail t-statistics as implemented in SPM12 (https://www.fil.ion.ucl.ac.uk/spm) and NiiStat (https://www.nitrc.org/projects/niistat). VLSM enables evaluation of the relationship of a predictor (e.g., DAZ of the meningioma) to a specific outcome (e.g., occurrence of MRS) at individual voxels.⁴⁰ Categorial variables (e.g., BINV, PTBE, WHO grade, histological subtype) were incorporated after dummy coding (0: reference group; 1: observational group). All voxels with signal were included in the analysis, resulting in 252 622 voxels analyzed. Threshold-free cluster enhancement⁴¹ was used to determine significant effects (H = 2, E = 0.5) as implemented by the Threshold-free Cluster Enhancement Toolbox in SPM12 (https://github.com/ChristianGaser/tfce). Multiple comparison correction was based on permutation testing with 5000 permutations.

Statistics

Statistical evaluation of clinical data was performed using SPSS (IBM SPSS Statistics for Windows, version 25.0, IBM Corp.) and custom-written MATLAB scripts including MATLAB statistics toolbox. For comparison of categorical data, we used a χ² test. Group comparisons of metric variables were based on nonparametric Kruskal-Wallis Tests. For multivariable analysis of multiple predictors on the occurrence of MRS, we applied a multivariable stepwise linear regression. P-values <.05 were considered significant. The results are expressed as mean ± SD.

Data availability

The data that supports the findings of this study are available on reasonable request.

Ethics Approval

The study was approved by the local ethics committee of the Medical Faculty of the Eberhard Karls University Tuebingen.

RESULTS

Patient Cohort

MRS occurred in 93/532 (17%) patients before surgery. In total, 44/93 (47%) showed a generalized onset. Within the focal onset cohort (44/93, 47%), most patients presented either motor (26/44, 61%) or somatosensory (5/44, 11%) symptoms; 13/44 (30%) patients presented focal onset seizures with cognitive symptoms. Overall, levetiracetam was the most common AED (68/532, 13%). Among 439 patients without MRS, 31 (7%) received prophylactic AEDs.

Spatial Distribution and Volumes of Meningiomas

Tumor masks were summed up to calculate spatial distribution of meningiomas. Lesion density maps showed meningioma hotspots in the middle one-third of the superior sagittal sinus, the falx, the frontoparietal transition zone (FPTZ, i.e., the coronal suture), and skull base structures such as sphenoid wing, and petroclival junction. There were only a few lesions observed at the parietooccipital convexities (Figure 1). The mean tumor volume was 35.2 ± 40.3 mL. Nonskull base meningiomas had significantly larger volumes (27.9 ± 34.7 mL and 40.5 ± 43.3 mL; Z = 4.59, P < .001).

Predictors of MRS

The univariable statistical analysis showed a significant association of MRS and male sex (χ² = 9.23; P = .002), atypical meningioma (χ² = 18.87; P < .001), WHO2 grade (χ² = 30.80; P < .001), BINV (χ² = 38.93; P < .001), PTBE (χ² = 25.51; P < .001), larger tumor volume (Z = 4.30; P < .001), and nonskull base location (χ² = 39.31; P < .001). Interestingly, the meningothelial meningioma subtype was negatively associated with seizure occurrence (χ² = 7.21; P = .007). However, meningothelial meningioma were significantly more like to occur at a skull base location (χ² = 43.38; P < .001). These results are summarized in Table 1.

TABLE 1.

Data Overview

	MRS+ N = 93	MRS− N = 439
Age	57.1 ± 13.6	56.3 ± 12.6	Z = 0.50; P = .615
Sex (%)
Male	37 (40)	107 (24)	χ² = 9.23; P = .002
Female	56 (60)	332 (76)
Histology (%)
Meningothelial	33 (36)	223 (51)	χ² = 7.21; P = .007
Fibroblastic	5 (5)	37 (8)	χ² = 0.98; P = .321
Transitional	10 (11)	51 (12)	χ² = 0.06; P = .812
Psammomatous	1 (1)	15 (3)	χ² = 1.44; P = .328
Angiomatous	4 (4)	13 (3)	χ² = 0.46; P = .515
Microcytic	5 (5)	9 (2)	χ² = 3.31; P = .079
Secretory	2 (2)	17 (4)	χ² = 0.66; P = .550
Lymphoplasmacyte	0 (0)	0 (0)
Metaplastic	1 (1)	4 (1)	χ² = 0.02; P = 1.000
Chordoid	2 (2)	8 (2)	χ² = 0.05; P = .689
Clear cell	0 (0)	0 (0)
Papillary	0 (0)	0 (0)
Rhaboid	0 (0)	2 (1)	χ² = 0.43; P = 1.000
Atypical	30 (32)	60 (14)	χ² = 18.87; P < .001
Anaplastic	0 (0)	0 (0)
WHO (%)
Grade 1	51 (55)	358 (81)	χ² = 30.80; P < .001
Grade 2	42 (45)	81 (19)
BINV (%)	31 (33)	40 (9)	χ² = 38.93; P < .001
PTBE (%)	64 (69)	175 (40)	χ² = 26.00; P < .001
Tumor volume	43.0 ± 39.0	33.1 ± 39.8	Z = 4.30; P < .001
Location (%)
Nonskull base
Parasagittal	26 (28)	62 (14)	χ² = 10.64; P < .001
Falcine	8 (9)	27 (6)	χ² = 0.75; P = .386
Cerebral convexity	34 (37)	75 (17)	χ² = 17.87; P < .001
Skull base
Cavernous sinus	1 (1)	12 (3)	χ² = 0.89; P = .483
Sphenoid winga	8 (9)	37 (8)	χ² = 0.003; P = 1.000
Clinoidal	6 (7)	43 (10)	χ² = 1.03; P = .311
Tuberculum sellae	0 (0)	28 (6)	χ² = 6.26; P = .008
Planum sphenoidale	0 (0)	10 (2)	χ² = 2.16; P = .223
Olfactory groove	3 (3)	17 (4)	χ² = 0.09; P = 1.000
Petroclival	0 (0)	31 (7)	χ² = 6.97; P = .003
CPA	0 (0)	42 (10)	χ² = 9.67; P < .001
Cerebellar convexity	0 (0)	9 (2)	χ² = 1.03; P = .311
Tentorial	2 (2)	19 (4)	χ² = 0.96; P = .556
Foramen jugulare	0 (0)	9 (2)	χ² = 1.94; P = .371
Foramen magnum	0 (0)	7 (2)	χ² = 1.50; P = .612
Temporobasal	3 (3)	3 (1)	χ² = 4.45; P = .035
Frontobasal	2 (2)	7 (2)	χ² = 0.14; P = .660
Seizure types (%)
Generalized onset	44 (47)
Focal onset	44 (47)
Motor	26 (28)
(Somato-)sensory	5 (5)
Cognitive	13 (14)
Unknown onset	5 (5)
Antiepileptic medication (%)
None	14 (15)	408 (93)	χ² = 238.83; P < .001
Levetiracetam	52 (56)	16 (4)	χ² = 185.74; P < .001
Valproate	4 (4)	1 (0)	χ² = 9.65; P = .002
Gabapentin	2 (2)	2 (1)	χ² = 1.12; P = .290
Carbamazepine	3 (3)	3 (1)	χ² = 2.46; P = .117
Oxcarbazepin	3 (3)	0 (0)	χ² = 9.07; P < .001
Phenytoin	7 (8)	3 (1)	χ² = 15.95; P < .001
Lamotrigine	4 (4)	1 (0)	χ² = 9.65; P = .002
Other	4 (4)	4 (1)	χ² = 3.89; P = .049
Postoperative complications (%)
Sensorimotor deficits	16 (17)	41 (9)	χ² = 4.96; P = .040
Ischemia	1 (1)	2 (1)	χ² = 0.52; P = .439
Bleedings	13 (14)	42 (10)	χ² = 1.61; P = .259
Infections	3 (3)	4 (1)	χ² = 3.17; P = .106

BINV, brain invasion; CPA, cerebellopontine angle; MRS, meningioma related seizures; PTBE, peritumoral brain edema; WHO, World Health Organization.

^aMedial/lateral sphenoid wing.

Patients were grouped based on the presence (MRS+) or absence (MRS−) of preoperative MRS. Statistically significant results are highlighted in bold.

As the univariate association of MRS with atypical meningioma, World Health Organization grade 2 (WHO2) grade and BINV might be biased by an interdependence of these factors (e.g., BINV is itself a grading criterion for WHO2 grade), we applied a multivariable stepwise linear regression analysis considering only variables with an independent explanatory contribution. Here, we examined the predictors of MRS (n = 534) depending on the 34 covariables (i.e., age, sex, histology subtype, WHO grading, BINV, PTBE, tumor volume, tumor location, see Table 1). After 5 iterations, a significant model (R = 0.385; F = 18.34, P < .001) was detected with BINV (B = 0.22, 95% CI [0.123, 0.310]; T = 4.56, P < .001), PTBE (B = 0.12, 95% CI [0.052, 0.177]; T = 3.61, P < .001) as well as the cerebral convexity (B = 0.16, 95% CI [0.084, 0.242]; T = 4.07, P < .001), the parasagittal (B = 0.15, 95% CI [0.066, 0.236]; T = 3.50, P < .001) and temporobasal (B = 0.37, 95% CI [0.089, 0.660]; T = 2.58, P = .010) location as significant positive predictors. Tumor volume (B = −0.04; T = −0.77, P = .443) and WHO grading (B = 0.02; T = 0.47, P = .642) had no significant effect. Importantly, WHO2 meningiomas had significantly larger tumor volumes than WHO1 tumors (28.4 ± 34.0 mL and 56.5 ± 49.1 mL; Z = 7.25, P < .001).

Voxel-Based Mapping of MRS Predictors

VLSM displayed a significantly higher MRS risk for meningiomas located in the FPTZ (Figure 2). The semiology of MRS depended on the meningioma DAZ. Focal onset seizures with motor and/or somatosensory symptoms were observed in perirolandic meningiomas. Opercular meningiomas were associated with focal onset seizures with cognitive symptoms. By contrast, frontal meningiomas led to generalized seizures (Figure 2).

FIGURE 2.

Spatial pattern of MRS. A, Frequency map shows the spatial distribution of meningiomas associated with MRS (n = 534). B, Voxel-based lesion mapping confirmed a significant higher incidence of MRS at the frontoparietal transition zone. Significant Z-scores are thresholded and corrected for multiple comparisons (P < .05; threshold-free cluster enhancement). Semiology of MRS depended on the meningioma DAZ. C, Focal onset seizures with motor and/or somatosensory symptoms were observed in perirolandic meningiomas. D, Opercular meningiomas were associated with focal onset seizures with cognitive symptoms. E, In contrast, frontal meningiomas led to generalized seizures. F, Multivariable voxel-based lesion mapping confirmed the spatial pattern of MRS semiology. Significant voxel (corrected P < .05; threshold-free cluster enhancement) are shown and color-coded according to the associated semiology. MRS, meningioma-related seizures.

Notably, MRS occurrence overlapped with areas of WHO2 meningiomas, BINV, and PTBE (Figure 3). To evaluate the spatial profile of the MRS predictors, we applied a multivariable VLSM to the occurrence of MRS considering the covariables BINV, PTBE, and their interaction (BINV × PTBE). This analysis indicated that BINV is the strongest predictor for the occurrence of MRS at the FPTZ. By contrast, PTBE is the main predictor for seizures associated with meningiomas of the frontal convexity and the sphenoid wing. A combination of these factors (BINV × PTBE) was the best predictor for MRS occurring in meningioma of the anterior frontal falx (Figure 3).

FIGURE 3.

VLSM of MRS, WHO2 meningioma (WHO2), BINV and PTBE. A, Univariable VLSM showed a significant predominance of WHO2 meningiomas, BINV and PTBE in overlapping areas covering the FPTZ and large areas of the frontal convexity. Significant Z-scores are thresholded and corrected for multiple comparisons (P < .05; threshold-free cluster enhancement). B, As previous analysis indicated BINV and PTBE as strongest predictors, we applied a multivariable VLSM to evaluate the spatial effect of BINV, PTBE and the interaction (BINV × PTBE) on the occurrence of preoperative MRS. VLSM indicated that BINV is the strongest predictor for MRS at the medial FPTZ. In contrast, PTBE is the main predictor for seizures associated with meningiomas of the lateral FPTZ, the frontal convexity and the sphenoid wing. Finally, the interaction of these factors (BINV × PTBE) was the best predictor for MRS occurring in meningioma of the anterior frontal falx. Significant Z-scores are thresholded and corrected for multiple comparisons (P < .05; threshold-free cluster enhancement). BINV, brain invasion; FPTZ, frontoparietal transition zone; MRS, meningioma-related seizures; PTBE, peritumoral brain edema; VLSM, voxel-based lesion symptom mapping; WHO2, World Health Organization grade 2.

Seizures After Surgical Resection

Gross total resection (i.e., Simpson grade<4) was achieved in 405/532 (76%). After resection, 31/532 (6%) patients experienced PostS > 3 months after surgery. The mean progression-free survival was 38 ± 35 months. Patients with preoperative MRS had a higher risk for postoperative sensorimotor deficits than patients without preoperative MRS (Table 1). Univariable statistics showed a significant association of PostS and atypical meningiomas (χ² = 8.07; P = .004), WHO2 grade (χ² = 4.50; P = .034), PTBE (χ² = 5.11; P = .024), larger tumor volume (Z = 3.02; P = .003), and nonskull base location (χ² = 18.36; P < .001). Notably, the meningothelial meningioma subtype was a negative predictor (χ² = 6.57; P = .010). The extent-of-resection (i.e., Simpson grade) had no effect on the occurrence of PostS. These results are summarized in Table 2.

TABLE 2.

Postoperative Seizures

	PostS+ N = 31	PostS− N = 501
Age	54.3 ± 15.5	56.6 ± 12.6	Z = −0.77; P = .439
Sex (%)
Male	12 (39)	132 (27)	χ² = 2.26; P = .133
Female	19 (61)	369 (74)
Histology (%)
Meningothelial	8 (26)	248 (50)	χ² = 6.57; P = .010
Fibroblastic	3 (10)	39 (8)	χ² = 0.14; P = .727
Transitional	5 (16)	56 (11)	χ² = 0.71; P = .384
Psammomatous	0 (0)	16 (3)	χ² = 1.02; P = .615
Angiomatous	2 (7)	15 (3)	χ² = 1.13; P = .260
Microcytic	1 (3)	13 (3)	χ² = 0.05; P = .831
Secretory	0 (0)	19 (4)	χ² = 1.22; P = .618
Lymphoplasmacytic	0 (0)	0 (0)	—
Metaplastic	1 (3)	4 (1)	χ² = 0.63; P = 1.000
Chordoid	0 (0)	10 (2)	χ² = 0.05; P = .689
Clear cell	0 (0)	0 (0)	—
Papillary	0 (0)	0 (0)	—
Rhabdoid	0 (0)	2 (1)	χ² = 0.12; P = 1.000
Atypic	11 (36)	79 (16)	χ² = 8.07; P = .004
Anaplastic	0 (0)	0 (0)	—
WHO (%)
Grade 1	19 (61)	390 (78)	χ² = 4.50; P = .034
Grade 2	12 (39)	111 (22)
BINV (%)	7 (23)	64 (13)	χ² = 2.43; P = .119
PTBE (%)	20 (65)	219 (44)	χ² = 5.11; P = .024
Tumor volume	56.6 ± 53.3	33.5 ± 38.5	Z = 3.02; P = .003
Location (%)
Nonskull base
Parasagittal	8 (26)	80 (16)	χ² = 2.05; P = .153
Falx	5 (16)	30 (6)	χ² = 4.89; P = .027
Cerebral convexity	12 (39)	97 (19)	χ² = 6.71; P = .010
Skull base
Cavernous sinus	1 (3)	12 (2)	χ² = 0.08; P = .546
Sphenoid winga	1 (3)	44 (9)	χ² = 1.16; P = .502
Clinoidal	3 (10)	46 (9)	χ² = 0.01; P = .758
Tuberculum sellae	0 (0)	28 (6)	χ² = 1.83; P = .397
Planum sphenoidale	0 (0)	10 (2)	χ² = 0.63; P = 1.000
Olfactory groove	0 (0)	20 (4)	χ² = 1.29; P = .622
Petroclival	0 (0)	31 (6)	χ² = 2.04; P = .245
CPA	0 (0)	42 (8)	χ² = 2.82; P = .160
Cerebellar convexity	0 (0)	9 (2)	χ² = 0.57; P = 1.000
Tentorial	1 (3)	20 (4)	χ² = 0.05; P = 1.000
Foramen jugulare	0 (0)	9 (2)	χ² = 0.57; P = 1.000
Foramen magnum	0 (0)	7 (1)	χ² = 0.44; P = 1.000
Temporobasal	0 (0)	6 (1)	χ² = 0.38; P = 1.000
Frontobasal	0 (0)	9 (2)	χ² = 0.57; P = 1.000
Simpson (%)
Grade 1	13 (42)	134 (27)	χ² = 3.37; P = .066
Grade 2	2 (7)	93 (19)	χ² = 2.92; P = .095
Grade 3	7 (23)	93 (19)	χ² = 0.31; P = .635
Grade >3	9 (29)	181 (36)	χ² = 0.64; P = .424

BINV, brain invasion; CPA, cerebellopontine angle; PostS, postoperative seizures; PTBE, peritumoral brain edema; WHO, World Health Organization.

^aMedial/lateral sphenoid wing.

Patients were grouped based on the presence (PostS+) or absence (PostS−) of postoperative MRS (PostS). Statistically significant results are highlighted in bold.

In addition, we applied a multivariable stepwise linear regression analysis to predict PostS depending on the covariables (i.e., age, sex, histology subtype, WHO grading, BINV, PTBE, tumor volume, tumor location, Simpson grading). After 2 iterations, a significant model (R = 0.165; F = 7.41, P < .001) was detected with tumor volume (B = 0.001, 95% CI [0, 0.001]; T = 2.82, P = .005) and the cerebral convexity location (B = 0.055, 95% CI [0.005, 0.104]; T = 2.18, P = .030) as significant positive predictors.

Considering the distinct impact of BINV on the occurrence of MRS, we hypothesized that BINV-associated MRS of the FPTZ might have a worse outcome after surgical resection than other seizures. In fact, BINV-associated MRS (n = 61, red area in Figure 4) had a higher risk for PostS (χ² = 7.01, P = .030). By contrast, PTBE-associated MRS (n = 161, green area in Figure 4) was not associated with sustained epilepsy after surgery (χ² = 0.30, P = .587). In line, VLSM detected a higher risk for PostS based on the location around the FPTZ (Figure 4).

FIGURE 4.

Voxel-based lesion symptom mapping of PostS after meningioma resection. A, Frequency maps indicated a higher risk of PostS in the parasagittal area. B, Voxel-based lesion symptom mapping confirms a significant higher risk of PostS for meningioma located at the perirolandic region and the adjacent parietal convexity. Significant Z-scores are thresholded and corrected for multiple comparisons (P < .05; threshold-free cluster enhancement). PostS, postoperative seizures.

DISCUSSION

In the past decades, there was extensive research to identify predictors of MRS providing conflicting results.^2-4,15 An underestimated factor in current research is tumor location. We hypothesized that an objective, assumption-free consideration of spatial tumor distribution (i.e., by VLSM) will help to resolve some contradictions and to provide a better understanding of MRS. This is, to the best of our knowledge, the first study to examine spatial features of MRS in relation to the DAZ applying the technique of VLSM. Our study suggests a significant association between meningiomas of the FPTZ and PreS as well as PostS. The importance of the FPTZ is highlighted by the predominance of preoperative focal seizures with sensorimotor symptoms in the present cohort. In addition, there was a higher risk of postoperative sensorimotor deficits after surgery in patients with preoperative MRS. Our study shows that BINV and PTBE are independent predictors of MRS. While BINV is strongly associated with seizures related to FPTZ meningiomas, PTBE is responsible for seizures in meningiomas of the frontal convexity and the sphenoid wing. BINV-associated MRS of the FPTZ have a worse outcome after surgical resection than other MRS.

PTBE, despite being closely related to BINV, is known to be a primary, independent risk factor for seizures.^11,42,43 Notably, PTBE may also develop in meningiomas without associated BINV, particularly by compression of cerebral veins or sinuses.^42,44-46 Specifically, sphenoid wing meningiomas and parasagittal meningiomas are known to cause PTBE due to their interactions with the middle cerebral veins and the sagittal sinus, respectively. PTBE increases cortical irritation through shifts in the extracellular ionic environment, altered neurotransmitter release, and blood-brain barrier breakdown, all of which contribute to a lower seizure threshold.^47-50 By contrast, BINV disrupts the pial-glial basement membrane,¹⁵ which triggers astrocytic activation and inflammation processes contributing to epileptogenesis in meningiomas.^51,52 To date, BINV is frequently considered a manifestation of high-grade meningiomas and is therefore used as an independent criterion for WHO grade 2 classification.^17,53 Independently, this study suggests that BINV is spatially related to the FPTZ, representing a novel observation not previously reported in the literature.

The foundation of BINV in meningiomas is multifactorial and not entirely understood. Genetic factors have been a primary focus in recent research. Higher grade (WHO grades 2 and 3) meningiomas of the falx cerebri and cerebral convexities are often related to pathogenic variants in the NF2 gene.^54-58 NF2 driver mutations, however, did not account for preoperative seizures; instead, their effect is presumed to be mediated through BINV.⁹ BINV has been linked to altered expression of beta-catenin and E-cadherin as members of the Wnt pathway,^59-61 a signaling cascade involved in epithelial-to-mesenchymal transition (EMT).⁶² EMT-like processes are considered to enable meningioma cells to infiltrate surrounding brain structures.^60,61 From an embryological perspective, the FPTZ (i.e., coronal suture and anterior fontanel) occupies a special position within the cerebral convexity. Here, neural-crest derived progenitor cells of the frontal meninges undergo EMT to contribute to cranial bone and suture formation.^63-65 Close reciprocal interactions of neural-crest derived progenitor cells with parietal mesodermal cells via paracrine signal pathways (such as Wnt) are known to regulate osteogenesis within the FPTZ.⁶⁵ This pathophysiological overlap of EMT processes in the development of the coronal suture and anterior fontanel and the invasion capacity of meningioma^60,61,65 might explain the predominance of meningiomas exerting BINV in the FPTZ. Besides, it is also possible that the FPTZ might have an intrinsic susceptibility for seizures.

Earlier work has indicated that seizure propensity varies across brain regions, with the frontal lobe showing an especially elevated likelihood of harboring epileptogenic zones.⁶⁶ Within this region, the premotor cortex has been highlighted as a site with heightened seizure vulnerability, particularly in patients with low-grade gliomas.⁶⁷ Our analysis in meningioma patients supports these previous studies. The mechanism underlying the seizure susceptibility of these areas has yet to be clarified.

Our results suggest that seizures related to PTBE of the frontal lobe might be reversible, while seizures related to meningiomas of the FPTZ have a significantly higher risk of persisting after tumor resection. This indicates a location-specific mechanism of PostS, which has not yet been described in previous studies. So far, there is evidence for an increased risk of PostS in high-grade meningiomas,^12,68-70 while the role of BINV remains unclear. As both WHO grading and BINV are unknown prior surgery, our findings have direct relevance for patient counseling and treatment decisions. Identifying the FPTZ as a location strongly associated with BINV and PreS as well as PostS highlights the importance of careful preoperative risk stratification. Patients with FPTZ meningiomas and BINV should be counseled about the elevated risk of seizures, which may lower threshold for initiating or continuing AED treatment. By contrast, seizures associated with PTBE, particularly in frontal convexity and sphenoid wing meningiomas, may carry a more favorable prognosis.

Limitations

There are some limitations of the study that should be addressed. First, our study is inherently limited by its retrospective nature. There was no systematic data on seizure frequency and burden (e.g., single vs multiple episodes). Furthermore, there were no data on the longitudinal development of PostS. The study uses MRI data acquired across different facilities, potentially introducing variability in lesion segmentation and interpretation of PTBE. In this study, PTBE was detected regardless of its size. However, PTBE volume may provide additional predictive value and should be addressed in future studies. Finally, there is increasing evidence that genomics and embryological features of meningiomas might affect WHO grading, BINV, and spatial profile.^19,21 This study, however, does not include genomic data. Combining VLSM approaches with genetic information may provide additional insights about the biology and pathology of meningiomas.

CONCLUSION

This study shows that MRS are significantly influenced by tumor location, BINV, and PTBE. Meningiomas invading the FPTZ have a higher risk of both preoperative and sustained PostS. PTBE contributes independently to MRS risk, particularly in frontal convexity and sphenoid wing tumors, but these seizures may be reversible after resection.

Key Points

• The FPTZ is the epicenter for meningiomas related to seizures and brain invasion.

• Epithelial-to-mesenchymal transition may explain epileptogenesis and invasiveness of meningioma at the FPTZ.

Importance of the Study

The importance of this study lies in its advancement of understanding MRS by integrating tumor location, BINV, and PTBE as critical factors. Using VLSM, the study precisely identifies spatial patterns in MRS-related, BINV-related, and PTBE-related meningiomas. The results emphasize the distinctive role of the FPTZ for epileptogenesis and invasiveness of meningioma. EMT is a critical embryological process in the formation of cranial structures at the FPTZ and may also contributes to the invasive behavior of meningiomas in this area. The study provides a framework for understanding the spatial and biological factors influencing MRS and highlights the potential for further research combining genomic, embryological, and VLSM approaches to refine our understanding of meningiomas and their impact on neurological outcomes.

Supplementary Material

neu-99-1047-s001.jpg ^{(168.3KB, jpg)}

Funding

This study did not receive any funding or financial support.

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

Supplemental Digital Content

Supplementary Figure 1. Flowchart study cohort selection. Flow diagram depicting selection of the study cohort. From 1175 patients identified in the institutional meningioma surgery database (2007-2017), those undergoing first-line surgery were retained (n = 1057), followed by patients with complete histopathological data (n = 1024) and complete clinical data, including preoperative and postoperative seizure information (n = 991). Cases with sufficient-quality preoperative MRI were then selected (n = 541). After excluding 9 intraventricular meningiomas, the final analytic cohort comprised 532 nonintraventricular tumors.