The course of tumor-related epilepsy in glioblastoma patients: A retrospective analysis

Purpose: Many patients with glioblastoma suffer from tumor-related seizures. However, there is limited data on the characteristics of tumor-related epilepsy achieving seizure freedom. The aim of this study was to characterize the course of epilepsy in patients with glioblastoma and the factors that influence it. Methods: We retrospectively analyzed the medical records of glioblastoma patients treated at the University Hospital Erlangen between 01/2006 and 01/2020. Results: In the final cohort of patients with glioblastoma (n = 520), 292 patients (56.2 %) suffered from tumor-related epilepsy (persons with epilepsy, PWE). Levetiracetam was the most commonly used first-line antiseizure medication (n = 245, 83.9 % of PWE). The onset of epilepsy was preoperative in 154/292 patients (52.7 %). 136 PWE (46.6 %) experienced only one single seizure while 27/292 PWE (9.2 %) developed drug-resistant epilepsy. Status epilepticus occurred in 48/292 patients (16.4 %). Early postoperative onset (within 30 days of surgery) of epilepsy and total gross resection (compared with debulking) were independently associated with a lower risk of further seizures. We did not detect dose-dependent pro-or antiseizure effects of radiochemotherapy. Conclusion: Tumor-related epilepsy occurred in more than 50% of our cohort, but drug-resistant epilepsy developed in less than 10% of cases. Epilepsy usually started before tumor surgery.


Introduction
The majority of patients with glioblastoma experience tumor-related seizures, either as a presenting symptom or during the course of the disease [1,2].It has been hypothesized that the development of epilepsy in individuals with glioblastoma depends on a combination of factors, including the inherent epileptogenic nature of tumor tissue, alterations in the tumor and surrounding microenvironment, and perturbations in the structure and function of neighboring brain regions [3].
Achieving durable seizure control is associated with reduced morbidity and improved quality of life and is an important treatment goal for patients with brain tumor-related epilepsy [4].While numerous studies have analyzed the role of epilepsy in tumor prognosis [5][6][7][8][9] less is known about the prognosis of this type of epilepsy in terms of seizure freedom.Previous studies investigating the prognosis of brain tumorrelated epilepsy have often been based on smaller, heterogeneous cohorts (low and high grade tumors) [2,[10][11][12].Seizure frequency is inversely correlated with tumor grade and growth rate [13][14][15][16][17].A recent meta-analysis evaluating the efficacy of antiseizure medication (ASM) in patients with grade II-IV gliomas found that seizure freedom rates at 12 months exceeded 70 % for levetiracetam (LEV) monotherapy [10].However, the characteristics that contribute to this favorable outcome and the factors that may lead to drug-resistant epilepsy in these patients are largely unknown.
We recently reported time-dependent risk factors for epileptic seizures in glioblastoma patients to address the question of how likely a patient is to have a seizure during a specific, defined interval of the glioblastoma disease course, namely preoperatively, early postoperatively, during or after radiochemotherapy.Using the same data set, this study aims to examine the course of epilepsy with seizure recurrence as the primary outcome parameter, while considering the time of first seizure as a covariate.In addition, we report on the occurrence of drugresistant epilepsy and identify factors that may influence the course of tumor epilepsy.We examined various tumor-related characteristics, seizure patterns, and treatment efficacy to gain insight into the clinical interplay between tumor, oncologic treatment, and epilepsy.

Methods
We performed a retrospective analysis of the electronic medical records of patients with glioblastoma treated at our institution.The digital patient files included admissions and outpatient follow-up appointments.All consecutive cases of newly diagnosed de novo glioblastoma seen between January 2006 and January 2020 were included in this study.Exclusion criteria were: (a) patients younger than 18 years at diagnosis, (b) individuals with a history of epilepsy, and (c) those with infratentorial or extracranial tumors.The study adhered to the STROBE guidelines and was approved by the Institutional Ethics Committee of the University of Erlangen-Nürnberg (No. 390_20Bc).Records were analyzed for demographic information, Karnofsky Performance Scale (KPS) at admission, tumor characteristics, oncological treatment, seizure occurrence, seizure semiology and ASM treatment.Gross total resection was defined as the absence of tumor on postoperative MRI scans.Subtotal resection, or partial resection, was defined as the presence of tumor on postoperative imaging.The tumor volume was calculated on preoperative T1-weighted contrast-enhanced MRI using the ellipsoid volume formula: π/6*D1*D2*D3, where D1, D2, and D3 correspond to the largest diameter of the compartment measured in three-dimensional plans (axial, sagittal, and coronal reformations).In patients with multifocal lesions, all lesions were summed for comparison [18][19][20].
The diagnosis of epilepsy was made according to the International League Against Epilepsy (ILAE) criteria (19), i.e. if the patient had a seizure related to the brain tumor at any time during the course of the disease.Both focal seizures and focal to bilateral tonic-clonic seizures were included.Seizures that occurred before the first surgery (i.e., the surgery that led to the tumor diagnosis) and had no apparent cause other than the underlying oncological disease were considered preoperative seizures.Seizures that occurred after surgery and before the start of radiotherapy (RT) were considered "early postoperative seizures".This definition was chosen in order to safely exclude acute symptomatic seizures after craniotomy, which are typically defined as seizures occurring within 7 days after surgery [21], from other seizures as our previous work demonstrated that acute symptomatic seizures do not increase the risk of recurrence [22].Seizures occurring during adjuvant RT or < 30 days after RT were classified as seizures during radiotherapy.Seizures occurring ≥ 30 days after primary oncological treatment (i.e., first surgery and adjuvant radiotherapy) were classified as posttherapeutic seizures.
Drug-resistant epilepsy was classified according to the ILAE definition in patients with epilepsy (PWE) who were not seizure-free after two ASMs [23].
To assess the risks associated with seizure recurrence, we compared patient characteristics of the group of PWE with a single documented seizure to those with more than one documented seizure (single versus recurrent seizures) using univariate and multivariate binary logistic regression analyses.
The multivariate analysis included significant clinical correlations and trends (p < 0.1) as well as oncologic parameters of first-line therapy (extent of surgery, radiation, chemotherapy) to assess their effect on seizure recurrence.We adjusted for age and follow-up time as potential confounders.In a second analysis, we analyzed the risks associated with drug-resistant epilepsy.
As this was an exploratory study, the significance level was set at p < 0.05 without correction for multiple testing.
Statistical analyses were performed using SPSS (version 26, SPSS Inc., IBM).Baseline characteristics of patients were expressed as percentage of patients, mean ± standard deviation (SD), median with interquartile range for data on overall survival, progression-free survival, and duration of follow-up.
We have previously published two analyses using this dataset investigating risk factors associated with seizures and status epilepticus [22,24].Any data not published in the article will be made available in anonymized form upon reasonable request by the corresponding author.

Antiseizure medication
Antiseizure medication (ASM) was started after the first seizure in the majority of patients with the exception of 13 (4.5 %) patients.Levetiracetam (LEV) was by far the most used drug initially (245 PWE, 83.9 %) with a median starting dosage of 1000 mg/d.Most PWE started with ASM monotherapy, with only 19/292 (6.5 %) receiving 2 ASM and 4/ 292 (1.3 %) receiving > 2 ASM right away.At the last follow-up, 117 PWE (40.1 %) were still taking their initial medication without any changes.26 patients (8.9 %) were not taking any ASM at the last followup.Dual-or polytherapy was established in 15 % of PWE at the last follow-up.ASM therapy was continued without changes in 117/292 (40.1 %) PWE.Table 2 shows the initial ASM and the ASM documented at the last follow-up (see Supplementary Table 2 for initial and last ASM in PWE with DRE).

Seizure freedom and drug-resistant epilepsy
136/292 PWE (46.6 %) remained seizure free after the first seizure throughout the follow-up period.Of the 150 patients with preoperative seizures, 19 had more than one seizure in the preoperative period.156 patients had more than one seizure, but only 27/292 (9.2 %) developed drug-resistant epilepsy according to ILAE criteria.Status epilepticus occurred in 48/292 PWE (16.4 %) and was the first manifestation of epilepsy in 18/292 PWE (6.2 %).

Risk factors for recurrent seizures after a first seizure
In univariate analysis, occipital (odds ratio [OR] 0.517, p = 0.05) or multifocal (OR 0.540, p = 0.03) tumor location and early postoperative seizures (OR 0.323, p = 0.02) were associated with a lower risk of seizure recurrence.Preoperative seizures (OR 1.622, p = 0.04), longer follow-up (OR 1.021, p = 0.01), and longer time between first seizure and ASM onset (OR 1.011, p = 0.075) were associated with an increased risk of seizure recurrence.
To address the issue of missing data on IDH mutation, we performed additional multivariate analyses including only patients with known negative IDH status (N = 344).Some results were no longer significant (such as follow-up and extent of resection), likely due to the reduced sample size leading to loss of power.However, there were no new or unexpected results.

Discussion
We retrospectively analyzed the course of tumor epilepsy in patients with glioblastoma and the factors that influence it.
Approximately 60 % of all glioblastoma patients treated in our institution suffered from tumor epilepsy.More than 80 % of all PWE were treated with LEV.LEV is one of the first line ASM in focal epilepsies and has demonstrated efficacy in brain tumor-related epilepsy [10,[25][26][27].Our study was not designed or powered to compare different ASMs.Valproic acid was rarely used, probably because of concerns about its rare side effect of causing bleeding problems during surgery [28], while benzodiazepines were relatively frequently used.
Over 50 % of all PWE had epilepsy onset prior to surgery, consistent with previous findings [1,6,9].Approximately 50 % of PWE experienced only a single seizure during the entire course of the disease, keeping in mind that most patients in our cohort were lost to follow-up within less than a year.Less than 10 % of PWE developed drug-resistant epilepsy (DRE) after failing 2 or more ASMs.In contrast, one third of the total population of PWE continue to have seizures despite ASM treatment, meeting the ILAE criteria for drug-resistant epilepsy [23,29].This confirmed the good prognosis of tumor epilepsy, in contrast to the poor prognosis of the underlying disease, as recently observed in a metaanalysis of different tumor types [10].However, data specifically addressing seizure freedom rates and DRE in glioblastoma patients are scarce; previous studies report seizure freedom rates of over 70 % at 12 months following surgery with DRE developing after surgery in up to 15 % of PWE [2,8,10,30].
A contributing factor to the good prognosis in terms of seizure control may be that nearly 50 % of the patients in our study underwent complete gross tumor resection, resulting in removal of the epileptogenic lesion.Compared to PWE with low-grade tumors, the time from first seizure to surgery is usually shorter in glioblastoma patients due to the characteristic presentation on MRI [31].In our cohort, the median time between surgery and first seizure was − 2.5 days.The lower survival rates and earlier resection in high-grade glioma patients may contribute to the lower seizure incidence and recurrence risk compared to low-grade glioma patients [1,13,[32][33][34].
In the assessment of risk factors for recurrent seizures, we included significant correlations from the univariate analyses in addition to potential confounders and parameters of primary oncologic therapy.We calculated another model using only significant correlations from the univariate analysis, which showed similar results (data not shown).We chose to include parameters of primary oncological therapy (surgery and radio-chemotherapy) because previous studies have found significant effects of radiation and chemotherapy on seizure freedom in lowgrade glioma [14,35,36].
Complete removal of the epileptogenic lesion by gross tumor resection correlates with a lower risk of seizure recurrence and thus better seizure control than debulking.If tumor tissue remains in situ, factors such as increased intracranial pressure, edema, hypoperfusion, neoangiogenesis, and structural and functional changes in the peritumoral tissue may continue to induce glioblastoma-associated epileptogenesis [37].Moreover, the inflammatory microenvironment in this region may also influence both epileptogenesis and tumor proliferation [38].As the risk of postoperative intratumoral bleeding or reactive peritumoral edema may be increased in patients who undergo debulking, this may explain our findings [39].In light of recent evidence of synaptic connections between neurons and brain tumor cells, suggesting a vicious cycle between brain tumor growth and epilepsy, it may even be possible that debulking creates more surface area for tumorimmune-neuron interaction, thereby increasing seizure activity [40].However, this is highly speculative.The non-significant effects for biopsy in this regard may well be explained by the small group sizes.When the first seizure occurred early postoperatively, the risk of recurrence was significantly lower than when the epilepsy started at a different time.This further supports the view that early postoperative seizures may not be associated with an increased risk of epilepsy because they can be considered acute symptomatic seizures which was already    hypothesized in previous work [21,22].Neither radiation nor chemotherapy had a significant effect on seizure control in our study.While beneficial effects of radiation on seizure control have been shown in low-grade gliomas [41], data in patients with glioblastoma are scarce.Climans et al. found a minimal effect of TMZ on seizure control in elderly patients with glioblastoma, which we did not reciprocate in our study [42].
The positive correlation between the length of follow-up and the likelihood of subsequent seizures is an expected finding, highlighting the importance of follow-up as a confounding factor in this type of investigation.In addition, our analysis revealed a statistically significant but small effect on seizure recurrence for patients with a longer interval between the first seizure and initiation of ASM treatment.Importantly, both associations had only small effect sizes (follow-up: OR 1.030, ASM initiation: OR 1.020).
Younger age and the occurrence status epilepticus (SE) as the first manifestation of epilepsy predicted drug-resistant tumor epilepsy (DRE).These results need cautious interpretation given the limited size of the study groups.However, SE was found to be a risk factor for drugresistant epilepsy before [43].It is left unclear from our results why younger age may be associated with DRE.Clearly, larger, prospective studies are needed to clarify this issue.

Limitations
Our study has several limitations, including its retrospective, monocentric design.
Accurate assessment of seizure frequency and semiology relied heavily on both patient and clinician reporting.Given the possibility that seizures may have been underreported or undocumented, there is a possibility that the achievement of seizure freedom may have been overestimated in our study.Because seizure frequency was inconsistently reported, especially in patients with frequent seizures (as opposed to first and second seizures), we focused on the analysis of patients with single versus multiple seizures.In particular, the incidence of drugresistant epilepsy may have been underestimated due to inconsistent documentation.We therefore restrained from analyzing the exact total number of seizures as these numbers may have been subject to bias due to the retrospective design.
Our results must be interpreted with caution because of the small number of patients in certain subgroups, particularly those with drugresistant epilepsy.Generalizability may be limited by the analysis of a monocentric convenience cohort.To address the potential bias caused by some patients receiving postoperative care at another hospital and being lost to follow-up, we included follow-up time as a covariate in each multivariate analysis.
Another major limitation of our study is the lack of missing neuropathological data.As MGMT promotor methylation analysis was not routinely performed in our hospital until recently, we lacked information on this important predictor of glioblastoma survival in > 60 % of the cases.As many patients treated before 2021 were included, we relied on the initial diagnosis given to the patients at the time of their disease and did not re-diagnose them as this would have been impossible in most patients.The new WHO classification no longer classifies astrocytoma with IDH1-mutation as glioblastoma and includes additional molecular markers which were not determined in most of our patients [44].In approximately 30 % of our cases, data on IDH1 status were missing.While previous studies in low-grade gliomas have found that seizures are associated with IDH mutations, the relationship in glioblastoma patients is less clear, with conflicting results from other studies [45].In our study, IDH1 mutation was not a significant factor for seizure recurrence.Although we did not find significant differences in epilepsy between patients with missing and present IDH1 mutation data (data not shown), we cannot exclude the possibility of bias caused by missing data.

Conclusions
In this monocentric, retrospective study, around 50 % of the glioblastoma patients developed the epilepsy preoperatively and around 50 % experienced only one seizure throughout the disease course.Drugresistant epilepsy was present in less than 10 % of cases.These findings demonstrate a generally good prognosis of tumor-related epilepsy which aligns with previous findings.Early postoperative seizures as first seizures and gross total resection were independently associated with a lower risk of further seizures.
Our findings may be valuable in clinical settings to assess the risk of seizure recurrence after the first seizure and identify patients at higher risk for such recurrences.Our results may aid in designing research that bridges the clinical course of tumor epilepsy and the intricate molecular patterns of the tumors involved.

Disclosure of conflicts of interest
H.M. Hamer has served on the scientific advisory boards of Arvelle, Bial, Corlieve, Eisai, GW, Novartis, Sandoz, UCB Pharma and Zogenix.He has served on the speakers' bureaus of or received unrestricted grants from Amgen, Ad-Tech, Alnylam, Bracco, Desitin, Eisai, GW, Nihon Kohden, Novartis, Pfizer, and UCB Pharma.The remaining authors have no conflicts of interest.

Ethical publication statement
We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.This study was approved by the ethics committee of FAU Erlangen-Nürnberg.

Fig. 1 .
Fig. 1.Flow chart of the study cohort.Patients with glioblastoma (total cohort), after application of the exclusion criteria (final cohort), and glioblastoma patients suffering from epilepsy (tumor epilepsy).

Fig. 2 .
Fig. 2. Kaplan-Meier Curve for absence of epilepsy.The plot illustrates the probability for the absence of epilepsy for the entire cohort (N = 520).The xaxis represents the time elapsed between the first diagnosis (=initial surgery, red vertical line) and the occurrence of the first seizure.Patients were censored if they either reached end of follow-up or died without experiencing a seizure.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)