If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Beijing Neurosurgical Institute, Capital Medical University, Beijing 100050, ChinaBeijing Tiantan Hospital, Capital Medical University, Beijing 100050, ChinaDepartment of Clinical Oncology, Beijing Academy of Critical Illness in Brain, Beijing 100069, China
Preoperative seizures were correlated to better survival outcomes in adult diffuse glioma.
•
In addition to early detection, IDH1 mutation might also be one of the major mechanisms.
•
This is the first study which provides a comprehensive standardized assessment on this issue.
Abstract
Purpose
Seizures are the most common presenting sign of patients with diffuse glioma. In the current study, we performed a meta-analysis to determine the correlation of seizures at presentation to survival outcomes in adult diffuse glioma, and the possible mechanisms were also discussed.
Methods
A comprehensive literature search was performed in PUBMED, EMBASE, Web of Science and the Cochrane Central Register of Controlled Trials. The pooled hazard ratio (HR) and corresponding 95% confidence interval (CI) were used to estimate effects. Heterogeneity among studies and publication bias were also evaluated.
Results
11 studies with 2088 patients were finally included for the current meta-analysis. Seizure-free preoperatively was significantly associated with a poor overall survival in patients with diffuse glioma, the pooled HR was 1.73 (95% CI 1.43–2.08, Z = 5.71, p < 0.001). Subgroup analysis was also performed by tumor grade, the same association was identified in both low-grade glioma (pooled HR 2.49, 95% CI 1.47–4.20, Z = 3.40, p < 0.001) and glioblastoma (pooled HR 1.46, 95% CI 1.27–1.68, Z = 5.24, p < 0.001). A significant correlation of seizure-free with a poor progression-free survival was also identified (pooled HR 1.42, 95% CI 1.06–1.92, Z = 2.33, p = 0.02), although only 3 studies comprising 368 patients were included.
Conclusion
The current study determined that seizures at presentation were an independent predictor of better survival outcomes in adult diffuse glioma. It is the first study which provides a comprehensive standardized assessment of the association between seizures at presentation with long-term survival outcomes in patients with diffuse glioma.
Epileptic seizures are the most common clinical manifestation of primary brain tumors, and the incidence of tumor-related seizures varies from 30 to 100% according to tumor type [
]. Among those tumors that can lead to epileptic seizures, glioma, accounting for over 80% of primary malignant brain tumors, is one of the most common types (the other one is glioneuronal tumor) [
]. Glioma can be divided into 4 grades (from grade I to grade IV) based on pathological evaluation according to the World Health Organization (WHO) classification of tumors of the central nervous system [
]. As WHO grade I gliomas (such as pilocytic astrocytomas) are stable and can be cured by surgery, few attentions have been paid to them. By contrast, the association between seizures and diffuse gliomas deserves more concerns.
According the 2016 WHO classification, the diffuse glioma refers to the WHO grade II and grade III astrocytic tumors, oligodendrogliomas (also includes grade II and grade III oligoastrocytomas diagnosed according to the 2007 WHO classification) and the grade IV glioblastomas (GBM) [
]. Low-grade gliomas (LGG, WHO grade II) are considered to be associated with a higher seizure incidence, seizures can be observed in 65–90% of LGG patients either as the initial symptom or as a symptom occurs during tumor progression [
]. Seizures carry both heavy financial and emotional burden on glioma patients, and significantly decrease their life quality: firstly, the effect of conventional treatment is not satisfied, despite combining antitumor therapies (such as surgical resection) with antiepileptic drugs, refractory epilepsy still exists in 20–30% of patients with glioma-related seizures [
]; secondly, epileptic seizures are arbitrary and unpredictable, a recent Mayo Clinic study shows that the unpredictable nature of seizures makes the patients loss their independent living ability, and results in great mental pressure to them [
]; thirdly, patients generally consider that seizures are closely related to the progression or recurrence of their tumors, which also adds to their anxiety [
Interestingly, according to our clinical experience, to glioma patients, although epileptic seizures mean a decreased life quality, they are often correlated with better survival outcomes. Several previous studies also reported the prognostic value of preoperative seizures in either LGG or HGG patients [
]. However, after a cursory review, we found that the studies investigating the association between seizure history and survival in glioma were much less than we anticipated, and the results were not so consistent. For instance, two recent studies showed that seizures at presentation were not an independent prognostic factor of survival in glioma patients [
]. This made us realized that it was necessary to perform a comprehensive review of the literature on this issue and give a definite conclusion on it. In the current study, we performed a meta-analysis to determine the correlation of seizures at presentation to survival outcomes in adult diffuse glioma, and the possible mechanisms were also discussed.
2. Methods
2.1 Search strategy and selection criteria
The current systematic review and meta-analysis was performed following the Preferred Reporting Items for Systematic Review and Meta-Analyses of individual participant data (PRISMA-IPD) guidelines [
]. We searched PUBMED, EMBASE, Web of Science and the Cochrane Central Register of Controlled Trials (CENTRAL) for relevant studies published from their inception to December 18, 2017. The search strategy was a combination of the following key words: “seizure” or “epilepsy” or “convuls*”, “glioma” or “astrocytoma” or “oligodendroglioma” or “oligoastrocytoma*” or “glioblastoma”, “survival” or “follow-up” or “prognos*” or “predict*” and “cohort” or “longitudinal” or “case-control” or “cross-sectional”. The search results were imported into Endnote X7 software (Thomson Reuters, New York, NY, USA) and duplicates were removed. Subsequently, titles and abstracts of remaining retrieved articles were further screened, studies would be excluded if they met any of the following criteria: 1) review articles, guidelines, or classifications; 2) case reports or small case series (cases <5); 3) meeting abstracts or abstract-only studies; 4) studies in children only; 5) studies only referred to other brain tumors (including WHO grade I glioma); 6) in vitro and animal experiment studies; 7) studies which investigated the prognosis of the seizures but not the survival of patients; 8) other irrelevant studies.
Next, potentially relevant full-text articles were obtained and re-screened. Studies which fulfilled the following inclusion criteria would be finally included in the current meta-analysis: 1) articles published in English; 2) reports of patients with preoperative glioma-related seizures; 3) data for progression-free survival (PFS) and/or overall survival (OS) were provided; 4) univariate and multivariate Cox Regression analysis were calculated and seizures at presentation was one of the initial enrolled factors; 5) multivariate-adjusted hazard ratio (HR) and corresponding 95% confidence interval (CI) were provided, or could be calculated. Reports from the same medical unit were examined carefully, if the results were reported by the same investigators and obtained on the same cohort, only the largest series would be included in the analysis.
2.2 Data extraction and management
The systematic literature search was conducted by a separate investigator, then two investigators screened the retrieved records and extracted data from all eligible studies independently. A standardized data collection form was used to record data from the eligible studies. Any possible conflicts would be resolved by group discussion with a third investigator.
2.3 Quality assessment
The 9-star Newcastle-Ottawa Scale (NOS, http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm) was applied to evaluate the methodologic quality of each eligible study. The overall quality of evidence was assessed by GRADE profiler software (version 3.6.1, McMaster University, Hamilton, Ontario, Canada) according to the grades of recommendation assessment, development and evaluation (GRADE) system [
]. The overall quality of literature was rated as high, moderate, low, or very low considering risk of bias, inconsistency, indirectness, imprecision, and publication bias.
2.4 Statistical analysis
Analyses were undertaken using Review Manager software (RevMan, version 5.3, Nordic Cochrane Centre, Cochrane Collaboration, Copenhagen, Denmark) and Stata software (version 14.0, Stata corporation, College station, Texas, USA). The HR and corresponding 95% CI were used as the primary measure to assess the correlation of preoperative seizures to survival in diffuse glioma. The significance was measured by the Z-test and p-value, a two-tailed p-value <0.05 was accepted as statistical significance criterion. Cochran's Q test, I2 statistic and H statistic were performed to evaluate the between-study heterogeneity. In the presence of significant heterogeneity (I2 > 50%; p-value of Cochran’s Q test <0.10 and I2 > 25%; H > 1.5 or 1.5 ≥ H ≥ 1.2 and the corresponding 95% CI does not include 1), data would be pooled using a random-effects model, and Galbraith plot should be carried out, otherwise, a fixed-effect model would be used. Moreover, sensitivity analysis should also be performed if the heterogeneity was significant.
A funnel plot was generated to evaluate the publication bias. Furthermore, Egger’s linear regression test and Begg's adjusted-rank correlation test were also performed to assess the publication bias, a two-tailed p-value <0.05 was considered statistically significant. If publication bias was suspected, the trim and fill method would be used to estimate the effects of the “missing” (unpublished) studies.
3. Results
3.1 Results of literature search
Fig. 1 shows the flowchart of our study. 493 publications from PUBMED, 279 from EMBASE, 168 from Web of Science and 94 from CENTRAL were identified through a comprehensive literature search. After the removal of 232 duplicate records, titles and abstracts of 802 references were reviewed and screened. 69 articles were selected for full-text assessment. 11 studies complied with the inclusion criteria were eventually included in the meta-analysis [
]. In addition to one prospective study, the remaining 10 studies were retrospective. The characteristics of the included studies are exhibited in Table 1.
Table 1Characteristics of studies included in the meta-analysis.
In the process of meta-analysis, ln(HR) and standard error (SE) needed to be calculated by HR and 95% CI, then imported into the Revman software. HR and 95% CI on the forest plot were exported according to the imported ln(HR) and SE. Accordingly, there will be slight differences between HR and 95% CI on the forest plot and those in the original article.
In the process of meta-analysis, ln(HR) and standard error (SE) needed to be calculated by HR and 95% CI, then imported into the Revman software. HR and 95% CI on the forest plot were exported according to the imported ln(HR) and SE. Accordingly, there will be slight differences between HR and 95% CI on the forest plot and those in the original article.
In the process of meta-analysis, ln(HR) and standard error (SE) needed to be calculated by HR and 95% CI, then imported into the Revman software. HR and 95% CI on the forest plot were exported according to the imported ln(HR) and SE. Accordingly, there will be slight differences between HR and 95% CI on the forest plot and those in the original article.
In the process of meta-analysis, ln(HR) and standard error (SE) needed to be calculated by HR and 95% CI, then imported into the Revman software. HR and 95% CI on the forest plot were exported according to the imported ln(HR) and SE. Accordingly, there will be slight differences between HR and 95% CI on the forest plot and those in the original article.
The study by Jiang et al. initially included 416 GBM patients while the study by Wang et al. initially included 211 LGG patients. However, as Jiang et al. introduced propensity score matching and Wang et al. had patients who lost follow-up, only 198 GBM and 150 LGG patients were enrolled in their final multivariate survival analysis.
The study by Jiang et al. initially included 416 GBM patients while the study by Wang et al. initially included 211 LGG patients. However, as Jiang et al. introduced propensity score matching and Wang et al. had patients who lost follow-up, only 198 GBM and 150 LGG patients were enrolled in their final multivariate survival analysis.
a In the process of meta-analysis, ln(HR) and standard error (SE) needed to be calculated by HR and 95% CI, then imported into the Revman software. HR and 95% CI on the forest plot were exported according to the imported ln(HR) and SE. Accordingly, there will be slight differences between HR and 95% CI on the forest plot and those in the original article.
b The study by Jiang et al. initially included 416 GBM patients while the study by Wang et al. initially included 211 LGG patients. However, as Jiang et al. introduced propensity score matching and Wang et al. had patients who lost follow-up, only 198 GBM and 150 LGG patients were enrolled in their final multivariate survival analysis.
The 11 selected studies comprised 2088 patients, with sample sizes ranging from 20 to 647 patients (mean 190, SD 171). All those studies enrolled 100 patients or more except the study by Pouratian et al. [
], and 3 studies (27.3%) included over 200 patients. 5 studies (45.5%) investigated LGGs only, while 5 studies (45.5%) included only GBMs, one study included both LGGs and anaplastic gliomas (AG, WHO grade III). Overall, 532 patients (25.5%) with LGG, 87 patients (4.2%) with AG and 1469 patients (70.3%) with GBM were analyzed. Among the 2088 patients enrolled in the current meta-analysis, seizures at presentation were observed in over 688 patients, while over 1052 patients were seizure-free at presentation (the accurate number of patients with or without preoperative seizures could not be obtained from the studies by Jiang et al. and Wang et al. [
The quality evaluation of each included study was performed according to the NOS (Table 1). The overall strength of evidence was evaluated with GRADE criteria and was rated as low.
3.3 Seizures at presentation and OS
Among 11 eligible studies, most of them (81.8%) reported a significant result that seizure-free was a detrimental prognostic factor for OS of glioma patients (HR ranged from 1.333 to 44.740). The pooled HR was 1.73 (95% CI 1.43–2.08, Z = 5.71, p < 0.001; Fig. 2), indicating that seizure-free was significantly associated with a poor OS. As a medium heterogeneity was observed (I2 = 47%; p = 0.04 for Cochran’s Q test; H = 1.4, 95% CI 1.0–2.0), a random-effects model was used. Galbraith plot were subsequently used to investigate the potential source of between-study heterogeneity. As shown in Fig. 3, the study by Pouratian et al. is the only outlier in the Galbraith plot under the dominant model. After removing this study, the heterogeneity decreased effectively (I2 = 22%; p = 0.25 for Cochran's Q test; H = 1.1, 95% CI 1.0–1.6).
Fig. 2The forest plot shows that seizure-free is significantly associated with a poor overall survival.
As LGG-related seizures and HGG-related seizures are quite different in mechanisms and epidemiology, we also conducted a subgroup analysis by tumor grade. The study by Quon et al. was not included in the analysis because it contained both LGGs and AGs. A significant association between seizure-free and a poor OS was identified in both LGG group (pooled HR 2.49, 95% CI 1.47–4.20, Z = 3.40, p < 0.001, Fig. 4A) and GBM group (pooled HR 1.46, 95% CI 1.27–1.68, Z = 5.24, p < 0.001, Fig. 4A). A medium between-study heterogeneity was observed within LGG group (I2 = 50%, p = 0.09 for Cochran's Q test) while no significant heterogeneity was observed within GBM group (I2 = 0%, p = 0.61 for Cochran's Q test). Additionally, significant heterogeneity was found between the two subgroups (I2 = 73.0%, p = 0.05 for Cochran’s Q test). After removing the study by Pouratian et al. (I2 decreased to 15% in LGG group, Fig. 4B), the heterogeneity between subgroups was still significant (I2 = 74.7%, p = 0.05 for Cochran's Q test, Fig. 4B), indicating that such heterogeneity might be another potential source of overall heterogeneity.
Fig. 4(A) The forest plot shows that seizure-free is significantly associated with a poor overall survival in both low-grade glioma group and glioblastoma group, a medium heterogeneity between subgroups is exist; (B) The forest plot after removing the study by Pouratian et al., the heterogeneity between subgroups is still significant.
Only 3 studies comprising 368 patients evaluated the association between preoperative seizures and PFS. A significant correlation of seizure-free with a poor PFS was also identified (pooled HR 1.42, 95% CI 1.06–1.92, Z = 2.33, p = 0.02, Fig. 5), and no significant heterogeneity was found (I2 = 34%, p = 0.22 for Cochran's Q test; Fig. 5).
Fig. 5The forest plot shows that seizure-free is significantly associated with a poor progression-free survival.
The funnel plot was used to evaluate the publication bias and asymmetry was observed, which suggested a potential publication bias (Fig. 6). Additionally, the Begg’s adjusted-rank correlation test (Z = 2.02, p = 0.043) and the Egger’s linear regression test (t = 3.05, p = 0.014, 95% CI 0.51–3.40; Fig. 7) were also performed and the results showed the existence of publication bias.
Fig. 6The funnel plot suggests a potential publication bias.
Fig. 8 shows a funnel-plot with imputed studies using the trim-and-fill method. 3 “missing” studies were indicated, with those studies, seizure-free was still associated with a poor OS (HR 1.63, 95% CI 1.45–1.83, p < 0.001, with the fixed model; HR 1.73, 95% CI 1.43–2.09, p < 0.001, with the random model). Including missing studies did not markedly affect estimates, which proved that our results were stable and credible.
Fig. 8The funnel plot with imputed studies using the Duval and Tweedie’s trim-and-fill method, shows the actual studies (circles) and the “missing” studies (circles with a square outside).
The “leave-one-out” scheme was adopted to assess the influence of studies on the pooled results. As shown in Fig. 9, the HR ranged from 1.67 to 1.82, the lower limit CI was 1.36 and the upper limit CI was 2.23. Such results suggested that the pooled HR was not materially altered, our results were robust.
Fig. 9Sensitivity analysis of the meta-analysis (for overall survival).
The current study determined that seizures at presentation were an independent predictor of better survival outcomes in adult diffuse glioma. To the best of our knowledge, this meta-analysis is the first study which provides a comprehensive standardized assessment of the association between seizures at presentation with long-term survival outcomes in patients with diffuse glioma.
The prognostic value of seizures at presentation in glioma was first reported early in 1980 by Scott et al. [
], they found that glioma patients with epilepsy preoperatively had a better survival than those without. To date, over 20 studies have reported the relationship between preoperative seizures and survival outcomes in diffuse glioma, and most of them showed that seizures were associated with a better survival [
Spontaneous and therapeutic prognostic factors in adult hemispheric World Health Organization Grade II gliomas: a series of 1097 cases: clinical article.
]. However, data in majority of those studies were insufficient for meta-analysis. In the current study, we combined the results from 11 eligible individual studies with 2088 glioma patients and indicated that seizures at presentation were a significant predictor of better OS. As the between-study heterogeneity was significant, Galbraith plot was subsequently used to spot the outliers, and the study by Pouratian et al. was identified as the potential source of the heterogeneity [
]. We inferred that was mainly caused by its small sample size (only 20 LGG cases).
Among 11 eligible studies, 5 studies investigated LGG-related seizures only while 5 studies simply analyzed GBM-related seizures. It has been widely recognized that seizure characteristics of LGG and GBM are quite different. For instance, the incidence of seizures in LGG patients varies between 65% and 90%, and seizure represents as the initial symptom in 70–90% of cases. The age at presentation is between 30 and 45 years. Seizures are mostly secondarily generalized (in 67% of LGG patients) [
]. As for GBM, the seizure frequency varies between 30% and 62%, and seizure represents as the initial symptom in about two thirds of patients. The average age at presentation is 60 years. 38% of GBM patients have focal seizures, 40% have focal seizures with secondary generalization and 12% have status epilepticus [
]. Such inconsistency of seizure characteristics may due to the differences in the mechanisms of epileptogenesis. The slow growth of LGG may result in the changes of peritumoral microenvironment, which can lead to epileptogenesis, while GBM is thought to induce epilepsy through abrupt ways of tissue damage [
]. Accordingly, we also performed a subgroup analysis and confirmed that the association between seizures at presentation and OS was still significant in both subgroups. The heterogeneity between the two subgroups was significant, in our opinion, such heterogeneity was also a manifestation of the differences between LGG-related and GBM-related seizures. Moreover, it was another potential reason of the significant overall heterogeneity.
Additionally, we also investigated the correlation of preoperative seizures with PFS, and found that seizures at presentation were also a significant predictor of better PFS. However, as the included studies number (3) was limited, this result need to be interpreted with caution.
The mechanisms behind the correlation of preoperative seizures to better survival outcomes are believed to be multifactorial and still far from clear. A generally agreed reason is that glioma patients presenting with seizures could be diagnosed at an earlier stage when the tumor is still small, so that there is a higher chance of total resection. However, as previously reported, the improved survival of patients with preoperative epilepsy was independent of the age, sex, tumor location, received treatments as well as tumor size [
]. In addition to early detection, there must be other mechanisms which contribute to the better survival of glioma patients presenting with seizures.
Isocitrate dehydrogenase 1 (IDH1) mutation is an early event in gliomagenesis. It has been identified as one of the most important molecular biomarkers for diffuse glioma classification and prognosis in the 2016 WHO classification [
]. IDH1 mutant glioma cells could reduce a-ketoglutarate to D-2-hydroxyglutarate (D2HG), and D2HG is structurally similar to glutamate, which is the most widespread excitatory neurotransmitter in our central nervous system. IDH1 mutant glioma cells can expose neurons to D2HG, disrupt the excitation-inhibition balance, and then induce seizures [
]. We therefore considered that IDH1 mutation might be one of the major mechanisms behind the correlation of preoperative seizures with better survival outcomes in gliomas. Additionally, there may be other specific yet unknown mechanisms, which need further investigation.
The current study has its clinical significance. Due to its dismal prognosis, diffuse glioma has always been a great challenge for neurosurgeons and oncologists. Although considerable progress in diagnosis and treatment has been made in the past decades, the prognosis of diffuse glioma, particularly GBM, is still poor [
Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial.
]. Pathological factors, including tumor grade, histological type, and molecular pathological changes such as IDH1 mutation, 1p19q co-deletion, telomerase reverse transcriptase (TERT) promoter mutation and α-thalassemia mental retardation X-linked protein (ATRX) mutation, have been reported as prognostic indicators of survival outcomes in glioma patients [
]. However, all markers above need a surgery to obtain the specimen for pathological evaluation. It is still difficult to predict the survival outcomes before surgery. In our study, seizures at presentation could be identified as an independent predictor for better survival outcomes in diffuse glioma. Such result suggests that glioma patients who are seizure-free preoperatively are most likely to require and benefit from more aggressive treatment.
Our analysis has some limitations. First, the included studies number was limited, more high-quality studies on this issue are still needed. Second, most of the studies included in the current meta-analysis were retrospective, additional studies are needed for further verification. Third, as the study only included a small number of patients with AG, subgroup analysis was only performed in LGG and GBM, relevant studies are also required. Overall, glioma-related seizures are still poorly understood currently, more attention needs to be paid in the future.
Ethical standards
The current meta-analysis does not report original data from human or animal subjects. Please consult the original publications for information about ethical procedures.
Conflict of interest
The authors have no actual or potential conflicts of interest related to this manuscript.
References
van Breemen M.S.
Wilms E.B.
Vecht C.J.
Epilepsy in patients with brain tumours: epidemiology, mechanisms, and management.
Spontaneous and therapeutic prognostic factors in adult hemispheric World Health Organization Grade II gliomas: a series of 1097 cases: clinical article.
Effects of radiotherapy with concomitant and adjuvant temozolomide versus radiotherapy alone on survival in glioblastoma in a randomised phase III study: 5-year analysis of the EORTC-NCIC trial.
30177-8&id=gr1.jpg){kind=link}
30177-8&id=gr2.jpg){kind=link}
30177-8&id=gr3.jpg){kind=link}
30177-8&id=gr4.jpg){kind=link}
30177-8&id=gr5.jpg){kind=link}
30177-8&id=gr6.jpg){kind=link}
30177-8&id=gr7.jpg){kind=link}
30177-8&id=gr8.jpg){kind=link}
30177-8&id=gr9.jpg){kind=link}