We conducted an observational, retrospective, single-centre cohort study at a tertiary surgical neuro-oncological centre between January 2006 and December 2024 during the era of the standard radiochemotherapy protocol [26, 27]. The manuscript was prepared in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist.

All methods were carried out in accordance with relevant guidelines and regulations. All experimental protocols were approved by the human research institutional review board (IRB00011687 – Collège de Neurochirurgie). The study received required authorizations (IRB#1:2025/30) from the IRB00011687 – Collège de Neurochirurgie. Informed consent was obtained from all subjects and/or their legal guardian(s), according to French legislation (observational retrospective study).

During the study period, 1324 patients were treated for a newly diagnosed supratentorial glioblastoma, IDH-wildtype. We excluded 609 patients (46.0%) treated with biopsy only, 18 patients (1.4%) with diffuse astrocytic glioma, IDH-wildtype, with molecular features of glioblastoma, WHO grade IV [30] and 21 patients (1.4%) without epileptic status data available. A total of 676 patients undergoing first-line surgical resection were included: 257 patients (38.0%) with Carmustine wafer implantation (mean 8.0 ± 2.0 wafers; range, 4–16) and 419 patients (62.0%) without. Study flowchart is presented in Fig. 1.

Baseline characteristics are detailed in Table 1. At diagnosis, 244 patients (36.1%) presented with epilepsy. Preoperative epileptic seizure prevalence did not differ between patients with (n = 92/257, 35.8%) and without (n = 152/419, 36.3%) Carmustine wafer implantation (p = 0.483). Preoperative seizure control did not differ between patients with (n = 247/257, 96.1%) and without (n = 386/419, 92.1%) Carmustine wafer implantation (p = 0.070).

Early postoperative epileptic control status was available for all 676 patients. Among them, 244/676 patients (36.1%) had a history of epilepsy at diagnosis, 201/244 (82.4%) achieved seizure control in the early postoperative period and 43/244 (17.6%) had uncontrolled seizures.

In the 432/676 (63.9%) patients without epilepsy at diagnosis, 414/432 (95.8%), achieved seizure control in the early postoperative period and 18/432 (4.2%) had uncontrolled seizures. Finally, 61/676 patients (9.0%) had uncontrolled seizures in the early postoperative period.

Uncontrolled seizures during the early postoperative period did not differ between patients with (n = 17/257, 6.6%) and without (n = 44/419, 10.5%) Carmustine wafer implantation (p = 0.098).

Risk factors of uncontrolled epileptic seizures in the early postoperative period are detailed in Table 2. In multivariable analysis, preoperative uncontrolled seizure (aOR 76.9, 95%CI 34.5-187.7, p < 0.001) was independently associated with uncontrolled seizure in the early postoperative period. Carmustine wafer implantation was not significantly associated with uncontrolled seizures during this period (aOR 0.78, 95%CI 0.36–1.60, p = 0.496).

Seizure status during the first six months of adjuvant oncological treatment was available for 599 alive patients; 77 patients died within this interval. Among them, 221/599 patients (36.9%) had a history of epilepsy at diagnosis, and 180/221 (81.4%) achieved seizure control during the first six months of adjuvant oncological treatment; 41/221 (18.6%) had uncontrolled seizures. In the 378/599 (63.1%) patients without epilepsy at diagnosis, 346/378 (91.5%), achieved seizure control during the first six months of adjuvant oncological treatment and 32/378 (8.5%) had uncontrolled seizures. Overall, 73/599 patients (12.2%) experienced epileptic seizures during the first six months of adjuvant oncological treatment: 32 (43.8%) were new-onset seizures in patients without preoperative epilepsy, and 41 (56.2%) occurred in patients with previously controlled seizures. Among the 548 alive patients with early postoperative epileptic seizure control, 64/548 (11.7%) developed uncontrolled seizures during the first six months of adjuvant oncological treatment. Uncontrolled seizures during the first six months of adjuvant oncological treatment did not differ between patients with (n = 26/224, 11.6%) and without (n = 47/375, 12.5%) Carmustine wafer implantation (p = 0.797).

Risk factors of uncontrolled epileptic seizures during the first six months of adjuvant oncological treatment are detailed in Table 3. In multivariable analysis, a history of epileptic seizure at diagnosis (aOR 2.38, 95%CI 1.43–3.98, p < 0.001) was independently associated with uncontrolled epileptic seizures during the first six months of adjuvant oncological treatment. Carmustine wafer implantation was not significantly associated with uncontrolled seizures during this period (aOR 0.92, 95%CI 0.55–1.54, p = 0.761).

In this retrospective, single centre study, we evaluated the impact of Carmustine wafer implantation on epileptic seizure control during surgical resection of newly diagnosed glioblastomas, IDH-wildtype in adults. Our findings indicate that Carmustine wafer implantation: 1) did not worsen the early postoperative epileptic seizure control; 2) did not worsen seizure control during the first six months of adjuvant oncological treatment; and 3) did not improve epileptic seizure control in either the early postoperative period or during the first six-postoperative months of adjuvant oncological treatment. Furthermore, we identified preoperative uncontrolled seizures as an independent predictor of uncontrolled seizures in the early postoperative period, and epileptic seizure at diagnosis as independent predictor of uncontrolled seizures during the first six months of adjuvant oncological treatment.

We recently investigated the natural history of epileptic seizures in adult patients with glioblastomas, IDH-wildtype[20]. In a large cohort of 1,006 patients, the cumulative incidence of tumor-related epilepsy increased substantially over the disease course, rising from 26.6% at diagnosis to 51.8% at the end of life. Similarly, uncontrolled epileptic seizures increased from 20.1% at diagnosis to 41.1% at the end of life. Importantly, a greater extent of surgical resection was associated with improved seizure control. In addition, data from resections of temporal glioblastomas further suggest that, when feasible, supratotal resection may confer additional seizure benefit compared with standard gross total resection [31]. However, compared with lower-grade IDH-mutant diffuse gliomas, glioblastomas, IDH-wildtype exhibit lower intrinsic epileptogenicity, which likely attenuates the potential benefits of tumor-directed therapies on epileptic seizure outcomes [32, 33]. Thus, while the antiepileptic effect of resection is less pronounced in glioblastomas than in lower-grade diffuse gliomas, surgical extent remains a key determinant of seizure control[20]. Radiotherapy exerts a dual effect on seizure dynamics. During treatment, approximately 40–45% of patients with high-grade gliomas experience transient worsening [34], likely due to treatment-related edema and neuronal irritation. In contrast, long-term epileptic seizure reduction is frequently observed following radiotherapy, with one study reporting significant improvement in nearly three-quarters of diffuse glioma patients [35]. This suggests that radiotherapy may modulate epileptogenesis primarily through tumor control, although robust evidence specific to glioblastoma, IDH-wildtype remains scarce. Temozolomide appears to play a modest role in epileptic seizure control. In a randomized trial in elderly glioblastoma patients, seizure incidence did not differ significantly between radiotherapy alone and radiochemotherapy (24% vs. 30%), though a trend toward delayed seizure onset was noted [36]. This suggests that the antiepileptic effect of Temozolomide is poor in glioblastoma, IDH-wildtype.

Given the potential influence of oncological treatments on epileptic seizure control, we assessed for the first time the natural history of epileptic seizures in homogeneous cohort of patients with a glioblastoma, IDH-wildtype treated with surgical resection and Carmustine wafer implantation. Previous studies evaluating the safety of Carmustine wafer implantation as a first-line therapy consistently reported no significant increase in early postoperative seizures [10–13, 22–24], while two studies suggested an increased epileptic seizure risk in recurrent glioblastomas [25, 37]. Across several prospective and retrospective studies, findings regarding postoperative epileptic seizure prevalence have been variable. Valtonen et al. in 1997 reported no difference in early postoperative seizures incidence between high-grade glioma patients receiving Carmustine wafers and placebo (19% vs 13%) [23]. Westphal et al. in 2003, in a randomized, double-blind trial of 240 patients, observed a slightly lower early seizure incidence in the wafer implantation group (33.3% vs 37.5%), raising the possibility of a modest antiepileptic effect [6]. Sabel and Giese in 2008, reviewed 29 studies in both first-line and recurrent settings, reporting epileptic seizure rates ranging from 19% to 33%, with higher rates in recurrent cases but no consistent difference compared with controls [38]. Notably, these early investigations predated the era of the standard radiochemotherapy protocol [26].

In the standard radiochemotherapy era, larger retrospective cohorts have confirmed the absence of a clear association between Carmustine wafer implantation and epileptic seizure risk. Attenello et al. in 2008, reported comparable 3-month epileptic seizure rates in 1013 craniotomies, including 288 with Carmustine wafer implantation (14.6% vs. 15.7%) [10]. Menei et al. in 2010, observed early postoperative epileptic seizure rates at 4.8% in newly-diagnosed cases and at 10% in recurrent cases across 163 patients receiving Carmustine wafer implantation [11]. Dixit et al. in 2011, described postoperative epileptic seizure rates of 5–16% in patients receiving Carmustine wafer implantation with radiochemotherapy, comparable to rates from radiotherapy-only series [22]. Duntze et al. in 2013, found a low epileptic seizure incidence of 3.3% in a prospective multicenter cohort of 92 patients [24], while Aoki et al. in 2014, found a 25% rate in a phase I/II trial of 16 patients, consistent with outcomes in patients not receiving Carmustine wafer implantation [21]. Larger retrospective series further reinforced these observations. Pallud et al. in 2015, in 787 patients (354 with Carmustine wafer implantation), found no significant increase in early postoperative epileptic seizures [12]. Bettag et al. in 2021, reported a 13% incidence of early postoperative epileptic seizure in 54 patients, with no association with intraoperative ventricular opening [39]. Taken together, these data suggest that Carmustine wafer implantation does not substantially modify seizure incidence in high-grade glioma patients. However, most series combined grade III and grade IV gliomas and predated the 2016 WHO classification, which distinguished IDH-mutant astrocytomas from IDH-wildtype glioblastoma [40]. Since glioblastoma, IDH-wildtype is intrinsically less epileptogenic than astrocytomas, IDH-mutant, grade 4, historical epileptic seizure data must be interpreted cautiously in this molecularly defined subgroup.

Our study provides the first dedicated, longitudinal assessment of the impact of Carmustine wafer implantation on epileptic seizure outcomes in a large, homogeneous cohort of adults with newly diagnosed glioblastoma, IDH-wildtype, treated in the current radiochemotherapy era. By addressing a critical knowledge gap, we offer robust evidence that Carmustine wafer implantation neither exacerbates nor improves postoperative seizure control, thereby informing clinical decision-making regarding antiseizure medication management. These findings are particularly relevant given the paucity of data specific to this molecularly defined glioblastoma subtype, which is known for its distinct biological behavior and lower intrinsic epileptogenicity compared to lower-grade diffuse gliomas. Carmustine wafers can be safely integrated into the standard oncological protocol without necessitating adjustments to antiseizure medication strategies. Furthermore, by highlighting preoperative seizure control as a key predictor of postoperative outcomes, we underscore the importance of individualized epilepsy management in glioblastoma patients.

This study represents the largest and most contemporary single-center analysis focusing on the impact of Carmustine wafer implantation on epileptic seizure outcomes in adults with newly diagnosed glioblastoma, IDH-wildtype. By restricting our cohort to this molecularly defined subtype and standardizing treatment protocols according to current guidelines, we minimize confounding variables related to histomolecular heterogeneity and evolving therapeutic standards. Our findings are directly applicable to real-world neurosurgical oncology practice, where Carmustine wafers could be used as an adjunct to maximal safe resection and standard radiochemotherapy. The homogeneity of our population, reflecting modern diagnostic criteria and treatment paradigms, strengthens the external validity of our conclusions, particularly for centers adopting similar management strategies. These results suggest not to modify the usual antiseizure medication after Carmustine wafer implantation.

The present results should be interpreted with caution given the retrospective, single-centre, and observational study design. First, Carmustine wafer implantation was not randomly assigned but determined intraoperatively by the neurosurgeon, introducing a potential selection bias. Carmustine wafers use may have been influenced by unmeasured confounders such as tumor location, intraoperative findings, or neurosurgeon preference. Second, key molecular markers were not systematically collected including the MGMT promoter methylation status (non-routinely assessed for clinical purposes according to Association of French-speaking Neuro-oncologists guidelines), which could influence both tumor behavior and seizure outcomes. Third, the diagnosis and characterization of epileptic seizures relied solely on clinical documentation, without the systematic use of electroencephalography monitoring. This approach is prone to recall, recognition, and reporting biases, potentially leading to underestimation or misclassification of seizure events. Additionally, the absence of quantitative data on seizure frequency over time and the lack of analysis of steroid use, which may independently affect seizure control, further constrain the robustness of our findings. Finally, we acknowledge that subjective biases in patient and physician reporting, as well as variations in documentation practices, may have influenced the assessment of epilepsy and seizure control. To address these limitations, future prospective, multicenter studies incorporating standardized EEG monitoring, detailed molecular profiling, and rigorous seizure assessment are warranted to better define the impact of Carmustine wafer implantation on postoperative epileptic seizure control.