Predictors of Postoperative Seizure Outcome in Low Grade Glioma: From Volumetric Analysis to Molecular Stratification

The importance of the extent of resection (EOR) has been widely demonstrated as the main predictor for survival, nevertheless its effect on tumor related epilepsy is less investigated. A total of 155 patients were enrolled after a first-line surgery for supratentorial Diffuse Low Grade Gliomas (DLGGs). Postoperative seizure outcome was analyzed stratifying the results by tumor volumetric data and molecular markers according to 2016 WHO classification. Receiver operating characteristic (ROC) curves were computed to asses EOR, residual tumor volume, and ΔT2T1 MRI index (expressing the tumor growing pattern) corresponding to optimal seizure outcome. A total of 70.97% of patients were seizure-free 18 months after surgery. Better seizure outcome was observed in IDH1/2 mutated and 1p/19q codeleted subgroup. At multivariate analysis, age (p = 0.014), EOR (p = 0.030), ΔT2T1 MRI index (p = 0.016) resulted as independent predictors of postoperative seizure control. Optimal parameters to improve postoperative seizure outcome were EOR ≥ 85%, ΔT2T1 MRI index ≤ 18 cm3, residual tumor volume ≤ 15 cm3. This study confirms the role of EOR and tumor growing pattern on postoperative seizure outcome independently from the molecular class. Higher ΔT2T1 MRI index, representing the infiltrative component of the tumor, is associated with worse seizure outcome and strengthens the evidence of common pathogenic mechanisms underlying tumor growth and postoperative seizure outcome.


Introduction
Seizure is the most common onset symptom in patients with supratentorial diffuse low grade gliomas (DLGG), with a seizure frequency ranging from 60% to 90% [1,2]. Tumor-related epilepsy tends

Postoperative Seizure Outcome Analysis
Postoperative seizure outcome was proportionally similar at the four analyzed time points (4,8,12, and 18-month follow-up). By considering the entire cohort of 155 cases, postoperative seizure control at 18 months was as follows: 110 patients (70.97%) were classified as Engel Class IA

Postoperative Seizure Outcome Analysis
Postoperative seizure outcome was proportionally similar at the four analyzed time points (4,8,12, and 18-month follow-up). By considering the entire cohort of 155 cases, postoperative seizure control at 18 months was as follows: 110 patients (70.97%) were classified as Engel Class IA (completely seizure free), 16 patients (10.32%) as Engel Class IB-ID, 23 patients (14.84%) as Engel Class II-III, and six patients (3.87%) as Engel Class IV. The distribution, according to different postoperative seizure outcome (Engel Class IA vs. IB-IV), of EOR, residual tumor, molecular class, and preoperative ∆T2T1 MRI index, and intraoperative protocol are represented in Figure 2.   Figure 2. Patients of Engel Class IB-IV required changes in AEDs therapy to optimize seizure control after surgery. As shown by the 18-month postoperative follow-up evaluation, these therapeutic changes, however, failed to produce complete seizure freedom, and no other patients achieved Engel Class IA.
AEDs changes was usually done for patients belonging to Engel Class IA and IB. For all the others, we proceeded with changings and/or optimization of the pharmacological treatments (i.e., add on therapy, a new AED as monotherapy, increased posology). Specifically, further surgery, radiotherapy or chemotherapy were not considered, in patients not seizure free after surgery, based on the fact that all cases showed no signs of tumor progression within the follow-up considered.
At univariate analysis the following parameters were associated with postoperative seizure outcome: frequency of preoperative seizures; seizure-onset features; preoperative ΔT2T1 MRI index; molecular class; EOR; AEDs in mono-or poly-therapy; and postoperative residual tumor computed on T2-weighted images. Patients of Engel Class IB-IV required changes in AEDs therapy to optimize seizure control after surgery. As shown by the 18-month postoperative follow-up evaluation, these therapeutic changes, however, failed to produce complete seizure freedom, and no other patients achieved Engel Class IA.
AEDs changes was usually done for patients belonging to Engel Class IA and IB. For all the others, we proceeded with changings and/or optimization of the pharmacological treatments (i.e., add on therapy, a new AED as monotherapy, increased posology). Specifically, further surgery, radiotherapy or chemotherapy were not considered, in patients not seizure free after surgery, based on the fact that all cases showed no signs of tumor progression within the follow-up considered.
At univariate analysis the following parameters were associated with postoperative seizure outcome: frequency of preoperative seizures; seizure-onset features; preoperative ∆T2T1 MRI index; molecular class; EOR; AEDs in mono-or poly-therapy; and postoperative residual tumor computed on T2-weighted images.
In a multivariate analysis model which considered all these variables, only age (p = 0.014), ∆T2T1 MRI index (p = 0.016), and EOR (p = 0.030) were shown to be independent predictors of outcome (Table 3).

ROC Analysis
In order to determine statistical clinical useful postoperative seizure outcome cutoff predictive values for EOR, preoperative ∆T2T1 MRI index and residual tumor, a receiver operating characteristic (ROC) curve was computed based on Engel class, using a binary outcome (Engel Class IA versus IB-IV) ( Figure 3). For the preoperative ΔT2T1 MRI index, the threshold of 18 cm 3 corresponded to the point with the highest sensitivity (0.689) and specificity (0.855), with a resulting area under the curve of 0.813 (CI 95% 0.731-0.895) and a predictive accuracy of 82.65%.
Based on the preoperative ΔT2T1 MRI index an example of proliferative and diffusive DLGG are shown in Figures 4 and 5.  The optimal threshold corresponded to an EOR of 85%, which was the point with the highest sensitivity (0.764) and specificity (0.644), with a resulting area under the curve of 0.783 (CI 95% 0.700-0.865) and a predictive accuracy of 72.90%.
Regarding the residual tumor, the optimal threshold corresponded to 15 cm 3 , which was the point with the highest sensitivity (0.556) and specificity (0.809), with a resulting area under the curve of 0.753 (CI 95% 0.663-0.842) and a predictive accuracy of 73.55%.
For the preoperative ∆T2T1 MRI index, the threshold of 18 cm 3 corresponded to the point with the highest sensitivity (0.689) and specificity (0.855), with a resulting area under the curve of 0.813 (CI 95% 0.731-0.895) and a predictive accuracy of 82.65%.
Based on the preoperative ∆T2T1 MRI index an example of proliferative and diffusive DLGG are shown in Figures 4 and 5. for the preoperative ΔT2T1 MRI index, the threshold of 18 cm 3 corresponded to the point with the highest sensitivity (0.689) and specificity (0.855), with a resulting area under the curve of 0.813 (CI 95% 0.731-0.895) and a predictive accuracy of 82.65%; (C) regarding the residual tumor, the optimal threshold corresponded to 15 cm 3 , which was the point with the highest sensitivity (0.556) and specificity (0.809), with a resulting area under the curve of 0.753 (CI 95% 0.663-0.842) and a predictive accuracy of 73.55%.

Discussion
In this retrospective study, which included 155 adult patients DLGG with preoperative drug-resistant tumor related epilepsy, postoperative seizure outcome was analyzed stratifying the results by tumor volumetric data and molecular markers.

Discussion
In this retrospective study, which included 155 adult patients DLGG with preoperative drug-resistant tumor related epilepsy, postoperative seizure outcome was analyzed stratifying the results by tumor volumetric data and molecular markers.
This study showed the following: (1) 70.97% of epileptic DLGG patients were in Engel Class IA 18 months after surgery; (2) Improved postoperative seizure outcome can be expected for EOR ≥ 85%, residual tumor ≤ 15 cm 3 , and preoperative ∆T2T1 MRI index ≤ 18 cm 3 . (3) Tumor infiltration index, expressed by ∆T2T1 MRI index, represents a quantitative evaluation of the diffusive and infiltrative tumor component as predictor of postoperative seizure outcome. (4) IDH1/2 mutation may represent the prevalent epileptogenic mechanism in presence of higher ∆T2T1 MRI index and consequent lower EOR.

The Role of EOR
The treatment paradigm of DLGGs is based on the principle of the onco-functional balance, which implies maximization of EOR with preservation of quality of life.
Xu et al. showed the existence of an EOR threshold for long-term seizure freedom corresponding to an EOR > 80% [17].
In a subsequent larger multi-center investigation, methodologically well-designed, Still et al. demonstrated that postoperative seizure control was more likely when EOR was ≥91% and/or when residual tumor volume was ≤ 19 cc in supratentorial DLGG patients [16].
The results reported in these studies with a postoperative follow-up of 6 months were also confirmed in our study, however, with a follow-up period of 18 months. Our data showed an optimum EOR threshold of ≥ 85% and a residual tumor threshold of 15 cm 3 to be associated with a higher likelihood of postoperative seizure control at one year. The epileptogenic focus and the tumor are not always overlapped. Hippocampectomy and corticectomy combined with lesionectomy in patients with DLGG and intractable preoperative tumor related epilepsy have been shown to improve seizure control [24]. The use of intraoperative electrocorticography (ECoG) to identify epileptogenic areas, guide surgical strategy, and improve postoperative seizure control in patients with LGGs still remains inconclusive, mainly due to its low accuracy to detect distant epileptogenic focus and to follow the spreading of epileptic activity [25]. Future intraoperative prospective studies are required to combine the intraoperative use of direct electrical stimulation (DES) and ECoG in order to optimize the DLGG surgery not only in terms of EOR achievable but also for the postoperative seizure outcome. Intraoperative integration between DES and ECoG may allow a supratotal resection, beyond the radiological margins of the tumor, when functionally possible.
Furthermore, during the sprouting of tumor cells in normal tissue, glioma cells release a high level of glutamate into the extracellular space. As a consequence, an imbalance between inhibitory and excitatory mechanisms is induced, generating neuron cell death, and promoting the migration of tumor cells [6,27].
As previously described, the preoperative ∆T2T1 MRI index provides an imaging estimate of tumor growing pattern prevalence and the EOR achievable [12,15].
Kinetic analysis in low grade gliomas highlighted that tumor growth results from two main mechanisms: proliferation and diffusion [28,29].
The prevalence of proliferation growth leads to a bulky tumor with a regular shape, determining similar tumor volume in both post-contrast T1-weighted MRI and T2-weighted preoperative MRI sequences; whereas the prevalence of the diffusive growth generates the tumor infiltration along the white matter, resulting in a complex shape with digitations more visible on T2-weighted images and less amenable of an extensive tumor resection [28,29].
Higher level of preoperative ∆T2T1 MRI index thus represents the prevalence of the diffusive growing mechanism.
In this study, we confirmed the role played by the tumor growth pattern (proliferative versus diffusive), expressed by ∆T2T1 MRI index on postoperative seizure outcome [15].
As a new feature, we provided a predictive cut-off value by the receiver operating characteristic (ROC) analysis.
According to ROC analysis, the optimum ∆T2T1 MRI index threshold ≤ 18 cm 3 was associated with a higher likelihood of long-term seizure control. Otherwise patients with a ∆T2T1 MRI index >18 cm 3 had a higher likelihood of postoperative seizure persistence at 18 months.
Assuming that the preoperative ∆T2T1 MRI index higher than 18 cm 3 reflects the prevalence of the diffusive and infiltrative tumor component, it could constitute an indirect imaging evaluation of changes in peritumoral tissue induced by tumor growth.
When the diffusive mechanism is predominant, tumor infiltrates the functional area limiting the resection, thus ∆T2T1 MRI index may provide a potentially estimation of the epileptic network development, allowing the preoperative detection of patients at greater risk of postoperative seizure persistence.

It is a Matter of Interaction between EOR and Tumor Growth Pattern
The simultaneous role played by the EOR and the ∆T2T1 MRI index in postoperative seizure outcome could reflect the relationship between the extent of resection achievable and the tumor growing pattern. The ∆T2T1 MRI index is the prognostic preoperative index of EOR itself, while the EOR is inversely related to the tumor growth pattern, as demonstrated in our previous study [12,15].
We can thus assume that the less infiltrative the tumor growth pattern is, the better the chances of greater EOR are and, consequently, the better the postoperative seizure control is.
Regarding the molecular assessment, only a few investigations have focused on the role played by the molecular biomarkers on postoperative seizure outcome in DLGG patients, reaching divergent results [21,23].
In a recent investigation, Neal et al. found a strong relationship between the higher IDH1-R132H rates and a severe postoperative seizure outcome, although the contribution to tumor related epilepsy by IDH 1/2mutation is not clear [21].
In contrast, Zhong et al. reported no significant difference in IDH status and seizure outcomes in 222 patients with WHO grade II gliomas [23].
In our study, when comparing seizure outcomes in patients with oligodendrogliomas and with astrocytomas (based on the 2016 WHO update), the majority of patients with oligodendrogliomas were Engel Class IA in comparison with patients with astrocytoma IDH1/2 mutated or astrocytoma IDH1/2 wild type (81.82%, 68.89%, and 50%, respectively; p = 0.001). It is likely that LGGs, which show less infiltrative growth, as oligodendrogliomas, are less prone to modify the tumor microenvironment, comparing to infiltrative lesions. Thus, electrical signaling of peritumoral tissues may be less impaired in oligodendrogliomas [20].
As a further confirmation regarding the role of EOR on seizure outcome, the Cox analysis revealed that the tumor molecular class did not result as an independent predictor at multivariate analysis, suggesting that EOR and the peritumoral infiltrative component are more important in predicting the outcome of epilepsy. Considering that the ∆T2T1 MRI index represents the infiltrative component of the tumor, it could be an indirect index of changes in peritumoral tissue induced by tumor growth and infiltration [15,26,27]. This index may be considered as a measure of persistent epileptogenic process after surgery. It is interesting to note that patients with prevalence of infiltrative tumor growing pattern, which means ∆T2T1 MRI index higher than 18 cm 3 , had a worse seizure outcome (66% of patient in Engel Class IB-IV, Figure 1B) and expressed the IDH1-2 mutation in more than 90% of cases. Huberfeld et al., in 2016, explained the relationship between epilepsy in glioma and IDH1/2 expression.
The epileptic discharge and tumor proliferation can be traced back to an imbalance in glutamate transporters determining an increase in concentrations of extracellular glutamate [6]. The presence of IDH mutated cells can explain seizure persistence in patients with a reduced EOR and high ∆T2T1 MRI index in the IDH-mutated tumors.
This hypothesis may explain the negative epileptic outcome in patient with higher ∆T2T1 MRI index. Indeed, at multivariate analysis, the ∆T2T1 MRI index (p = 0.016) resulted as a stronger predictor of postoperative outcome in comparison with EOR (p = 0.030), underlying the importance of the infiltrative tumor component, which is generally not removed because it is functional, in seizure persistence. The availability of predictive factors for postoperative seizure outcome could provide a useful tool to guide therapeutic antiepileptic strategy after surgery, avoiding pharmacological overtreatment improving the patients' quality of life.

Limitation and Future Directions
There are several limitations of our study. The most important one is based on the retrospective nature of the investigation The peritumoral cortex may contain epileptogenic foci, which may directly affect postoperative seizure control [8,24,30]. Future studies are thus required to better investigate the correlation between supra-total resection, when functionally achievable, and seizure outcome.
Although in the present study gliomas were classified accordingly to the currently used 2016 WHO classification of the brain tumors, it is desirable in the future to molecularly characterize diffuse astrocytic gliomas, IDH-wildtype, in order to recognize those that can indeed be classified as diffuse astrocytic glioma, IDH-wildtype, with molecular features of glioblastoma, WHO grade IV [31,32]. Additional information deriving from next generation sequencing analysis would help stratify postoperative seizure outcome in DLGGs, using markers with known pathophysiological roles in epilepsy such as glutamate metabolism/clearance [5].
In the pattern of results reported, the cognitive statuses of the patients were not considered. The impact of tumor related epilepsy on the pre-surgical neuropsychological examination could prove to be useful to better assess the effects of tumor growth itself and the influence of tumor related epilepsy or medication on the cognitive status of the patients. It has been shown that DLGG patients can present emotional and personality changes in their post-surgery examination [33]. An important aspect would be analyzing the effects of tumor related epilepsy on the post-surgery quality of life of patients in terms of emotional processing and personality.
In closing, considering the lack of standardized protocol for tumor related epilepsy management, both before and after surgery, it should be important to plan a multidisciplinary approach considering the complex therapeutic profile of DLGGs patients [34]. In detail, a preoperative study as for epilepsy surgery with prolong Video-EEG recordings for patients with tumor related epilepsy characterized by complex semiology not directly associated with tumor location, could be useful in future studies to assess the spreading of epileptic discharges and plan the function possibility of resection and the intra-operative position of the strips.
A tailored AEDs treatment should be adopted for each patient, considering that changes in seizure type or worsening in Engel Class could be related to tumor progression. There are numerous points of interest to be noted: homogeneous data collection in a monoinstutional case-series; strict definition of postoperative seizure control as Engel Class A patients; postoperative seizure follow-up at 18 months after surgery concomitant with control MRI, to rule out cases with tumor progression; integration of tumor volumetric data and molecular data, according to 2016 WHO classification, to stratify postoperative seizure outcome; ROC analysis to determine EOR, residual tumor volume, and ∆T2T1 MRI index corresponding to optimal seizure outcome.

Study Population
A total of 155 adult patients with tumor related epilepsy underwent surgery at our institution for primary DLGG (January 2007-May 2018).
Seizures were the onset symptom and all patients developed seizure not fully controlled with medical treatment before surgery. At least two AEDs were given in successive monotherapies or together in politherapy, resulting in a drug-resistant epilepsy, according to the International League Against Epilepsy (ILAE) definition [35].
The 2017 ILAE classification was applied to classify the type of seizures [36]. Patients were enrolled according to the following criteria: age ≥ 18 years; preoperative MRI suggestive of supratentorial low grade glioma; no previous surgery, chemo-or radio-therapy; at least 18 months of follow up, with concomitant MRI control, to rule out cases with tumor progression. Objective evaluation of EOR preoperatively and postoperatively on MRI images in DICOM format based on T2-weighted MRI sequences; revision of histopathological specimens by using the new 2016 WHO Classification; diagnosis of drug-resistant tumor related epilepsy, according to the ILAE definition [37].
Patients were evaluated preoperatively, at discharge and, during the follow up, every 6 months. Engel Class at 18-month follow-up was used to compute predictors of postoperative seizure outcome [36].
No patient underwent adjuvant therapy with radiotherapy or chemotherapy during the period of postoperative follow-up.
The local Ethics Committee, Comitato Etico Unico Regionale del Friuli Venezia Giulia, approved this investigation (protocol N.0036567/P/GEN/EGAS, ID study 2540). Considering that the study was retrospective, written consent to participate in the study was not applicable. Written informed consent was obtained for surgery.

Surgical Procedure
All patients were surgically treated with the aim of the intraoperative brain mapping technique both at cortical and subcortical level [38].
The awake surgery protocol was selected following the standard protocol previously described [39].

Volumetric Analysis
Tumor volume data were obtained by analyzing structural imaging data routinely acquired during pre-surgery and post-surgical investigations in axial 3D T2-weighted and 3D post-contrast T1-weighted MRI slices. All tumor segmentations were realized by using the OSIRIX software tool (GNU LESSER, General Public License, Geneva, Switzerland) [40,41]. Specifically, the tumor growing pattern, expressed by MRI ∆T2T1 index, and EOR were computed as previously described [12,15].
Briefly, the tumor growing pattern and the EOR were assessed as listed: (1) preoperative tumor volume segmented on axial T2-weighted MRI images-preoperative tumor volume segmented on axial T1-weighted images. (2) (preoperative tumor volume-postoperative tumor volume)/(preoperative tumor volume) in axial T2-weighted MRI axial images.

Histological and Molecular Analysis
Histological and molecular data were retrospectively analyzed according to the 2016 WHO classification [42].
Immunohistochemistry (IHC) for Ki67 and IDH1R132H, fluorescence in situ hybridization (FISH) to evaluate 1p/19q codeletion and analysis of the genetic status of O6-methylguanine-DNA-methyltransferase (MGMT) promoter and isocitrate dehydrogenase (IDH1/2) genes were performed as previously described. Gliomas were classified as methylated when the average percentage of methylation of CpG islands was ≥ 8% [43].

Statistical Analysis
Categorical variables were reported as percentages, continuous variables were reported as mean ± standard deviation or median and range as appropriate, according with the data distribution. Normality of the continuous variables was tested using the Shapiro-Wilk test, the t-test or Mann-Whitney U-test as appropriate, was used to compare continuous variables between groups. For the outcome analysis, Engel classification was dichotomized as Class IA versus Class IB-IV (patients were either completely seizure free or not completely seizure free). In univariate analysis, the variables considered as possible prognostic factors were age, sex, preoperative tumor volume, tumor histological subtype, molecular markers, tumor side, preoperative seizures feature, seizure onset characteristics and frequency, time between seizure onset and surgery, intraoperative protocol used, EOR, residual tumor volume, and preoperative ∆T2T1 MRI index.
Multivariate stepwise backward analyses included all variables significant at p = 0.05 in univariate analysis. The results were presented as odds ratios and 95% confidence intervals.
For EOR, preoperative ∆T2T1 MRI index and residual tumor threshold, the cut-off values able to discriminate, with high sensitivity and specificity, the postoperative seizure control, was determined by De Long's nonparametric receiver operating characteristic (ROC) analysis with exact binomial estimation of confidence intervals (CI) of the area under the curve (AUC).
All analyses were conducted using Stata/SE (version 14.0 Stata Corp.) for Mac. All two-tailed statistical significance levels were set at p < 0.05.

Conclusions
The association between a worse seizure outcome and the prevalence of an infiltrating tumor growing pattern expressed by high ∆T2T1 MRI index reinforces evidence of common mechanisms underlying both epileptogenesis and tumor growth.
The individual evaluation of the tumor growing pattern and the estimation of the EOR may thus represent a helpful tool in the early identification of patients with an increased risk of seizure persistence after surgery.
Tumor volumetric information could be useful when deciding and discussing prognosis and potential postoperative seizure outcomes to better handle the entire management, starting from pre-to post surgery.
Considering the role of IDH 1/2 mutation in both tumor growing and epileptogenesis, and considering that up to 80% of DLGGs carried out this mutation [44], potential future goals could be represented in the development of anti-epileptic drugs targeting the underlying biochemical pathology related to the mutation of IDH 1/2.