Postsurgical Seizure Outcome for Epilepsy Patients According to Histopathological Diagnosis: A Single-Center Experience

Introduction

Epilepsy is a prevalent neurological disorder, affecting an estimated 10 million people in China.¹ It is characterized by recurrent, abnormal electrical discharges in the cerebral cortex, clinically manifesting as seizures.² These recurrent seizures not only impair patients’ physical and mental health but also impose a heavy financial burden on families and the healthcare system, consuming substantial societal resources.^3,4

Focal epilepsy results from abnormal electrical discharges that originate in specific, localized regions of the brain.⁵ Although seizures can be controlled with medications in most patients, more than one-third of individuals with focal epilepsy continue to experience recurrent seizures. The epileptogenic zones in these patients may display complex pathological features, such as structural lesions (eg hippocampal sclerosis or malformations of cortical development) or functional network reorganization.⁶ Recent studies suggested that sustained overactivation of neuroinflammatory pathways, such as the HMGB1-TLR4 axis, can perpetuate network hyperexcitability, resulting in persistent seizures.⁷ Furthermore, the dysfunction of functional proteins like the Sigma-1 receptor may lead to the collapse of intracellular homeostasis, thereby diminishing neuronal responsiveness to treatment.⁸ Epilepsy surgery is the most effective treatment to control seizures in these patients, further improving cognitive function, behavior, and quality of life.^9,10 In fact, approximately 65% of patients with focal epilepsy have been reported to achieve seizure freedom following surgery.¹⁰

Favorable surgical outcome is been linked to younger age at surgery, shorter epilepsy duration, and lower preoperative antiseizure-medication burden.^9,11,12 Recent evidence shows that the histopathology of the epileptogenic zone is closely associated with prognosis. However, studies on histopathology report wide variations in the rates of seizure freedom after surgery. In analyses focusing on focal cortical dysplasia (FCD), the proportions of patients with FCD achieving seizure freedom range from 8.3% to 86.4%.^13,14 Other research suggests that patients with FCD often fare worse than those with other pathologies.^10,15,16 While some MRI-diagnosed FCD cohorts report favorable outcome in up to ~70% of cases.¹⁷

Although the prognostic importance of histopathology in the epileptogenic zone is increasingly recognized, most prior studies have been constrained by small cohort sizes, limiting detailed histopathological subtyping. Moreover, most investigations have focused on only one pathology, rather than examining multiple coexisting histopathologies. In particular, the impact of dual or multiple pathologies on surgical outcome remains poorly explored.

Based on this, we assembled a large cohort of patients to ensure sufficient representation across all histopathological categories. We systematically analyzed their preoperative, surgical, and postoperative data to further explore the relationship between histopathology and postsurgical outcome.

Materials and Methods

Patient Selection

This study screened 974 epilepsy patients treated at the Department of Neurosurgery, First Affiliated Hospital of Fujian Medical University from 2017 to 2024. First, 243 patients lacking postoperative histopathological diagnoses were excluded. Then, 11 patients without presurgical MRI results were excluded. Additionally, 271 patients with specific histopathology diagnoses (such as malignant lymphoma, metastatic carcinoma, meningioma, high-grade neuroepithelial tumors, and melanoma) were removed. Ultimately, 449 patients met our inclusion criteria and were enrolled in the study (Figure 1).

Figure 1 Selection of study participants.

Data Collection

We systematically collected patient demographics (sex, age at seizure onset, age at surgery), clinical characteristics (epilepsy duration, number of preoperative ASMs), imaging and monitoring data (presurgical MRI, use of stereoelectroencephalography [sEEG]), surgical parameters (type and anatomical site of surgery, use and result of intraoperative electrocorticography [ECoG]), histopathological diagnosis, and seizure outcome (Engel classification). Surgical interventions were classified into two types: resective surgery and minimally invasive laser interstitial thermal therapy (LITT). We categorized MRI findings into four distinct groups: MRI+/pathology+: structural abnormality evident on both MRI and histopathology; MRI-/pathology+: no lesion on MRI, but structural abnormality confirmed by histopathology; MRI-/pathology-: no structural abnormality on either MRI or histopathology; MRI+/pathology-: lesion visible on MRI without a matching histopathological abnormality.

Histopathological Assessment

Postoperative histopathological diagnoses were classified into 10 categories: low-grade epilepsy-associated neuroepithelial tumor (LEAT), cerebral vascular malformations (including cavernous malformation, arteriovenous malformation, and venous malformation), malformations of cortical development, subdivided into focal cortical dysplasia (FCD) and other non-FCD cortical malformations (MCD-nFCD); nonspecific lesion (including gliosis and “no lesion”), hippocampal sclerosis (HS), dual or multiple pathologies (ie more than one coexisting pathology), and other rarer entities (such as extranodal sinusoidal histiocytosis, schwannoma, epidermoid cyst, pituitary adenoma, etc.). “no lesion” refer to the absence of discernible structural abnormalities on histopathologic examination.

Statistical Analysis

Baseline characteristics for patients achieving Engel I versus Engel II–IV at ≥12 months after surgery were compared using χ2 tests for categorical variables and the Mann–Whitney U-test for continuous variables. Categorical data are reported as proportions, and continuous variables as medians. The interactions and multicollinearity among relevant variables were examined. Univariate and multivariate logistic regression analyses were performed to estimate odds ratios (ORs) for seizure freedom (Engel I) versus recurrent seizures, stratified by histopathological diagnoses to evaluate how different subtypes influence prognosis. In addition, univariate and multivariate logistic regression models assessed other potential predictors of seizure outcome, including age at seizure onset, age at surgery, epilepsy duration, and the number of preoperative ASMs and MRI results. Apply propensity score matching to balance data imbalance in MRI results. Statistical analyses were conducted using IBM SPSS 24 and R version 4.4.2 through RStudio 2026.01.0, and p < 0.05 was considered statistically significant.

Results

Patient Characteristics

Of the 449 patients, 158 (35.2%) had epilepsy onset before 18 years of age, and 98 (21.8%) were under 18 at the age of surgery. Age at seizure onset ranged from less than 1 year to 75 years, while age at surgery spanned from 1 to 75 years. One hundred and seventy (37.9%) were female. The median epilepsy duration was 6 months (interquartile range [IQR]: 0.0007–480 months), with 155 (34.5%) patients having a duration of one year or less and 294 (65.5%) exceeding one year. The number of preoperative ASMs ranged from 0 to 14. The median follow-up duration was 47 months (IQR: 12–90 months). Demographic characteristics of the study patients were shown in Table 1.

Table 1 Demographic Characteristics of the Study Population (n = 449)

Surgical Results and Risk Factors

Of the 449 patients included, complete postoperative outcome data were available for 397 (88.4%). Among them, 15 patients had died, leaving 382 for further analysis. Their basic characteristics were shown in Table 2. Among these 382 patients, 292 (76.4%) achieved Engel I, while 90 (23.6%) fell into Engel II–IV categories. There were no significant differences between these two groups with respect to age at seizure onset, age at surgery, or surgical type. However, patients with recurrent seizures (Engel II–IV) had a significantly longer duration of epilepsy (median 36 months vs 3 months, p < 0.001) and were prescribed a greater number of preoperative ASMs (median 1 [range 0–14] vs.0 [range 0–7], p < 0.001) compared to seizure-free patients. Additionally, a higher proportion of recurrence cases had undergone sEEG monitoring (14.4% vs 6.5%, p = 0.027).

Table 2 Baseline Characteristics of Patients with Complete Data (N = 382)

Analysis of presurgical MRI findings revealed that the largest subgroup was MRI+/pathology+ (83.1%, n = 373). While the MRI-/pathology- group was the smallest (only 8 cases, 1.8%). MRI-/pathology+ was observed in 50 cases (11.1%) and MRI+/pathology- in 18 cases (4.0%). Notably, all patients with LEAT showed MRI+/pathology+. Among the MRI-/pathology+ group, FCD accounted for the largest proportion (72.0%, n = 36). Fisher’s exact test identified a significant difference in prognosis between the MRI+/pathology+ group and the MRI-/pathology+ group (p = 0.003). There were more patients with recurrent seizures among the MRI-/pathology+ group (20.0% vs 8.6%, p = 0.004), while Engel II–IV outcome was less common in the MRI+/pathology+ group (71.1% vs 87.0%, p = 0.001).

Given the potential complex relationship between MRI findings, preoperative ASMs, and epilepsy duration, we specifically explored interactions as well as multicollinearity among these three variables and a Mann–Whitney U-test between duration, ASMs, and MRI findings. The result shows that the interaction between MRI and ASMs (p = 0.842), the interaction between MRI and duration (p = 0.901), and the interaction between ASMs and epilepsy duration (p = 0.381) were all insignificant. In addition, the multicollinearity analysis shows that the VIF values of the three types of independent variables are all well below the critical value of 10 (MRI: 1.175, ASMs: 1.400, duration: 1.383), and the tolerances are all well above 0.1 (MRI: 0.851, ASMs: 0.714, duration: 0.723). There is no serious multicollinearity problem among these independent variables. However, the Mann–Whitney U-test revealed that MRI-negative patients generally had longer epilepsy duration (p < 0.001) and greater use of ASMs (p < 0.001), indicating substantial baseline confounding. To control for this imbalance and explore the independent effects of factors, we performed logistic regression analysis. Due to sample size imbalance in the MRI groups, we merged the two subgroups with negative postoperative histopathology (MRI-/pathology- group and MRI+/pathology- group).

Univariate logistic regression analysis was performed to assess the association between surgical outcome and the following variables: age at seizure onset, age at surgery, epilepsy duration, number of preoperative ASMs, intraoperative ECoG results, and MRI results. The analysis revealed that recurrent seizures after surgery were significantly associated with longer duration of epilepsy (OR = 1.006, 95% CI 1.004–1.009), a higher number of preoperative ASMs (OR = 1.357, 95% CI 1.172–1.572), and the result of MRI-/pathology+ (OR = 2.857, 95% CI 1.470–5.556). In multivariate logistic regression, both longer epilepsy duration (adjusted OR = 1.005, 95% CI 1.001–1.010) and a greater number of preoperative ASMs (adjusted OR = 1.235, 95% CI 1.026–1.487) remained independently associated with increased odds of seizure recurrence. MRI findings showed no independent predictive value (p = 0.724). The multicollinearity diagnosis shows that the variance inflation factors of all independent variables are well below 10 (all VIFs ≤ 3.461, all tolerances ≥ 0.289), indicating that there is no serious multicollinearity in the model and that the estimation results are reliable. Results of univariate and multivariate analyses were shown in Table 3.

Table 3 Uni- and Multivariate Regression Analysis About Basic Characteristics in Patients with Favorable vs Unfavorable Outcome (N = 382)

To solve the imbalance between the groups of MRI+/pathology+ and MRI-/pathology+, we conducted a 1:1 nearest neighbor matching method based on propensity scores. Using MRI grouping (MRI+/pathology+ vs MRI-/pathology+) as the treatment variable and key baseline covariates, including age at seizure onset, age at surgery, epilepsy duration, number of preoperative ASMs, and intraoperative ECoG results, we successfully constructed a baseline-balanced cohort (all test SMD > 0.1). The results show that the prognostic difference between MRI+/pathology+ and MRI-/pathology+ groups was no longer statistically significant (p = 0.262).

Surgery Characteristics

Most of the 449 patients (97.1%, n = 436) underwent resective surgery, including 24 patients who had prior sEEG monitoring. The remaining 13 patients were treated with MRI-guided LITT. Among the 292 patients with Engel I outcome, the proportion of resection was 97.6% (n = 285), and only 7 (2.4%) underwent LITT. In contrast, of the 90 patients with Engel II–IV outcome, resective surgery was performed in 85 cases (94.4%) and LITT in 5 cases (5.6%).

Postoperative Histopathological Diagnosis

The distribution of histopathological diagnoses among the 449 patients was as follows: LEAT 34.3% (n = 154); cerebral vascular malformations 24.5% (n = 110), including cerebral cavernous malformation (12.0%, n = 54) and other cerebral vascular malformations (12.5%, n = 56); malformations of cortical development 20.0% (n = 90), including FCD (18.9%, n = 85) and MCD-nFCD (1.1%, n = 5); nonspecific lesion 6.5% (n = 29), consisting of gliosis (5.3%, n = 24) and “no lesion” (1.1%, n = 5); HS 0.7% (n = 3); dual or multiple pathologies (6.0%, n = 27); and others rare entities 8.0% (n = 36). The two largest categories of histopathological diagnosis were LEAT and cerebral vascular malformations. LEAT had the highest rate of favorable outcome (Engel I, 88.4%). In contrast, FCD and gliosis were associated with substantially lower seizure-free rates. (59.2% and 58.8%, respectively). Notably, Engel II–IV patients had a significantly lower proportion of LEAT compared to patients with Engel I (16.7% vs 39.0%, p < 0.001) and a higher rate of malformations of cortical development (35.6% vs 16.8%, p < 0.001), particularly FCD (34.4% vs 15.4%, p < 0.001).

Univariate logistic regression analysis was conducted for the six main histopathological categories: LEAT, MCD, cerebral vascular malformations, nonspecific lesion, dual or multiple pathologies, and others. The analysis showed that LEAT showed a significant association with a reduced risk of postoperative seizure recurrence compared to the other five categories (OR = 0.312, 95% CI 0.171–0.570), whereas MCD was closely related to seizure recurrence (OR = 2.736, 95% CI 1.611–4.647). No statistically significant relationships were found for cerebral vascular malformations (p = 0.361, OR = 0.770, 95% CI 0.440–1.349), nonspecific lesion (p = 0.113, OR = 2.094, 95% CI 0.839–5.225), dual or multiple pathologies (p = 0.151, OR = 1.937, 95% CI 0.785–4.779), or others (p = 0.273, OR = 0.544, 95% CI 0.183–1.617). In multivariate logistic regression, MCD remained an independent predictor of seizure recurrence (adjusted OR = 4.136, 95% CI 2.122–8.063). Furthermore, nonspecific lesion (adjusted OR = 3.897, 95% CI 1.418–10.715) and dual or multiple pathologies (adjusted OR = 3.619, 95% CI 1.330–9.846) emerged as significant risk factors for recurrence. No independent associations were observed for cerebral vascular malformations (p = 0.185, OR = 1.603, 95% CI 0.797–3.224) or others (p = 0.872, OR = 1.101, 95% CI 0.341–3.557). Results of univariate and multivariate analyses were shown in Table 4. These findings indicate that MCD, nonspecific lesion, and dual or multiple pathologies are independent predictors for unfavorable postsurgical outcome.

Table 4 Uni- and Multivariate Regression Analysis About Histopathologic Results of Patients with Favorable vs Unfavorable Outcome (N = 382)

Univariate logistic regression was performed for six histopathological subtypes: cerebral cavernous malformation (CCM), other cerebral vascular malformations excluding CCM (VM-nCCM), FCD, MCD-nFCD, gliosis, and “no lesion”. The analysis indicated that FCD was associated with a significantly increased risk of postoperative seizure recurrence (OR = 2.884, 95% CI 1.683–4.941). By contrast, MCD-nFCD showed no significant association with recurrence (p = 0.850, OR = 0.809, 95% CI 0.089–7.332). This indicates that the association between MCD and poorer surgical outcome is primarily observed in its FCD subtype. No statistically significant associations were observed for CCM (p = 0.140, OR = 0.531, 95% CI 0.229–1.231), VM-nCCM (p = 0.792, OR = 1.095, 95% CI 0.556–2.157), gliosis (p = 0.088, OR = 2.378, 95% CI 0.878–6.442), or “no lesion” (p = 0.946, OR = 1.082, 95% CI 0.111–10.536). A multivariate logistic regression model was then constructed, including these six subtypes, plus the category of dual or multiple pathologies. In this model, FCD (OR = 4.290, 95% CI 2.258–8.151), gliosis (OR = 4.359, 95% CI 1.502–12.653), and dual or multiple pathologies (OR = 3.558, 95% CI 1.338–9.466) emerged as independent predictors of postoperative seizure recurrence. No significant association were found for CCM (p = 0.855, OR = 1.090, 95% CI 0.434–2.736), VM-nCCM (p = 0.064, OR = 2.076, 95% CI 0.959–4.494), MCD-nFCD (p = 0.698, OR = 1.557, 95% CI 0.166–14.581), and “no lesion” (p = 0.535, OR = 2.076, 95% CI 0.207–20.860). Results of univariate and multivariate analyses were shown in Table 5. These findings further indicate that FCD, gliosis, and dual or multiple pathologies are important determinants for unfavorable postsurgical outcome.

Table 5 Uni- and Multivariate Regression Analysis About Detailed Histopathologic Results in Patients with Favorable vs Unfavorable Outcome (N = 382)

Discussion

This study demonstrated that seizure outcome of focal epilepsy patients is highly associated with specific histopathological subtypes, with histopathological diagnosis serving as an important and independent predictor of surgical success. LEAT was related to a markedly higher probability of achieving seizure freedom. In contrast, MCD (especially FCD), nonspecial lesion, dual or multiple pathologies, and gliosis were independently associated with an increased risk of recurrent seizures.

In this study, 59.2% patients with FCD achieved Engel I, which aligns closely with a meta-analysis of 37 studies reporting an average seizure-freedom rate of 55.8% ± 16.2% in FCD cohorts.¹⁸ FCD location is known to influence lesion size, age at epilepsy onset, and surgical outcome.¹⁹ Presurgical Evaluation, such as MRI, may fail to localize the FCD in some patients, contributing to unfavorable surgical outcome. Recent research using graph neural networks has shown promise for detecting FCD in MRI-negative epilepsy patients, potentially leading to better outcome.²⁰

Focal seizure affects both local epileptogenic networks and distributed structural-functional network reorganization, characterized chiefly by reduced connectivity between the lesion and the rest of the brain.^21–23 Surgical disconnection may promote compensatory increase in postoperative network integration; However, persistently low integration can allow residual pathological network to reactivate (eg under stress or fatigue), leading to seizure recurrence.⁶ The functional MRI studies in MCD patients demonstrate altered functional network at the lesion, perilesional areas, and distant neocortex,^21,24 suggesting that impaired structural-functional connectivity may contribute to unfavorable surgical outcome in temporal MCD and FCD patients. New study demonstrates that clemastine, a non-selective σ1 receptor agonist, exhibits significant neuroprotective effects against PTZ-kindled seizures and cognitive deficits. Its anti-seizure and pro-cognitive actions are mediated by acting on σ1 receptors, highlighting the crucial role of σ1 receptors in improving seizures and protecting cognitive function.⁸ This discovery offers hope for treating epilepsy in patients with poor surgical prognosis.

Patients with gliosis had a lower chance of achieving seizure freedom. Studies reported that a well-defined aura and the disappearance of epileptiform discharges on intraoperative ECoG were predictive of favorable postoperative seizure outcome.²⁵ Conversely, absence of auras may indicate rapid propagation of epileptiform discharges to remote brain regions, hindering precise localization of the epileptogenic zone. Persistent frequent spikes on post-resection ECoG may suggest incomplete resection of the epileptogenic zone. Taken together, these factors may contribute to recurrent seizures after surgery.

This study specifically focused on dual or multiple pathologies, which were characterized by the concurrent presence of two or more lesions. In such cases, it is often unclear whether seizure originates from a single lesion or from multiple independent lesions. sEEG may help to identify the epileptogenic focus; However, if sEEG captures only a limited number of seizures, localization remains highly uncertain.²⁶ In patients with dual or multiple pathologies, separate lesions and adjacent tissue changes may reflect the sequential evolution of a single pathological process and thereby contribute to an expanded epileptogenic zone.²⁷ In addition, a recent study suggested that HMGB1 activates the TLR4/NF-κB signaling pathway and stimulates RAGE, causing hyperexcitability and acute or chronic epilepsies. This is mediated by glial cell activation and the release of proinflammatory cytokines, establishing a self-sustaining cycle of neuroinflammation.⁷ This may explain the pathological progression and expansion of the epileptogenic zone. These complexities can preclude complete resection of the epileptogenic focus and consequently contribute to unfavorable surgical outcome.²⁸ When technically and clinically feasible, simultaneous resection of all identifiable lesions is generally recommended for patients with dual or multiple pathologies.^29,30

This study found that preoperative MRI results are not an independent predictor of postoperative seizure recurrence in epilepsy patients with positive histopathology. MRI-negative and pathology-positive status is a marker of more severe epilepsy, characterized by longer duration and greater use of ASMs. Clinical prognosis assessment should prioritize these severity indicators over MRI findings alone. A study examining the prognosis of MRI-negative pediatric epilepsy patients demonstrated favorable postoperative outcome with the employment of video electroencephalography, magnetic resonance spectroscopy, and magnetoencephalography to localize the epileptogenic focus.³¹ In another study, MR-negative and positive FCD patients had a comparable surgical prognosis, emphasizing the importance of comprehensive presurgical evaluations.³² These suggest that with precise localization of the epileptogenic focus, MRI-negative/pathology-positive patients may achieve comparable favorable surgical outcome to MRI-positive patients.

Furthermore, the lack of significant prognostic differences between the MRI-/pathology- and MRI+/pathology- groups may be attributable to their relatively small sample sizes. This limitation may undermine the comprehensiveness of our analysis linking MRI-negative findings to prognosis, explaining discrepancies with prior studies. A recent study suggested that patients with an electroclinically identified focus achieve better surgical outcome compared to those patients with negative MRI findings and negative pathology results.³³ Future prospective studies with larger cohorts are needed, with particular consideration of pathology-negative patients’ prognoses.

Limitation

Several limitations of our study should be acknowledged. First, patients with isolated hippocampal sclerosis were too few for regression analysis, and FCD was not analyzed by subtype, which may have prognostic relevance. The prognosis of FCD patients needs more detailed discussion. Furthermore, due to the retrospective design, although we adjusted for key surgical and demographic confounders, other unmeasured factors that may affect surgical results, such as detailed nutritional status and socioeconomic determinants of health, were not available for analysis.³⁴ These factors should be further explored. Additionally, the retrospective design led to a notable imbalance in sample size between MRI groups. Despite applying multivariable logistic regression and propensity score matching, unmeasured confounding may persist.

Conclusion

This single-center retrospective cohort study explored the relationship between histopathology and surgical outcome in epilepsy patients. Specific histopathological categories, including FCD, gliosis, and dual or multiple pathologies, were significantly associated with seizure recurrence after surgery, while LEAT was linked to a lower risk of recurrence. Shorter duration of epilepsy and a lower number of presurgical ASMs were identified as independent predictors of favorable surgical outcome. However, the relationship between MRI findings and prognosis was not significant, suggesting that patients with negative MRI but positive pathology may achieve a favorable prognosis as MRI-positive patients.