PDF
Citation
Print
Introduction
Glioblastoma (GBM) is the most common malignant primary brain tumor in adults, with a poor survival rate. Maximal safe resection followed by standard combined chemoradiation with temozolomide (TMZ) has been recommended as the first-line treatment. However, the overall survival (OS) rate is less than 2 years, leading to the investigation of novel or adjunctive agents [1-3].
Seizures are known to occur in 30%-60% of patients with GBM, prompting hospital visits where antiepileptic drugs (AEDs) are commonly prescribed. Over the last decade, levetiracetam (LEV) has emerged as the preferred AED for patients with GBM due to its favorable safety profile and effectiveness. There is increasing evidence supporting the role of LEV in GBM development and progression [4-17]. Studies have indicated that LEV may boost the effectiveness of TMZ, enhancing its therapeutic action against GBM. Given these insights, the examination of LEV’s impact on the survival outcomes of patients with GBM is of paramount importance.
The Health Insurance Review and Assessment Service (HIRA) database is a nationally acknowledged resource that contains the medical data of approximately all South Koreans. It collects data from healthcare professionals and insurance claims and ensures reliability through unique patient IDs, enabling vast and long-term data tracking. The database uses a distinct coding system that enables the precise identification of medical treatments and ensures accurate comparisons and analysis. We conducted a large-scale nationwide population-based analysis using the HIRA database of South Korea. Our study included patients with GBM in the HIRA database to determine how the duration of LEV use affects the survival outcomes of patients with GBM.
Materials and Methods
1. Data collection
We analyzed data from the Korean HIRA database to evaluate the effect of LEV on the clinical outcomes of patients with GBM. The data of individuals diagnosed with the International Classification of Diseases (ICD)-10 code C71.x who underwent cranial surgery and tumor removal (codes: S4634, S4635, S4636, and S4637) were obtained. These patients received TMZ chemotherapy and radiation therapy within 6 weeks post-surgery, between January 2007 and November 2021. We excluded patients diagnosed under the age of 18 years, those with more than a 90-day gap between their initial diagnosis and surgical procedure, those who died within 90 days of surgery, and those diagnosed with an additional malignancy. The study included patients diagnosed between 2007 and 2018, with follow-up data collected until November 2021. Patients diagnosed after 2018 were excluded to ensure a sufficient follow-up period for reliable survival analysis, considering that the median survival time after GBM diagnosis is typically between 12 and 24 months. Patients were subsequently divided into two categories based on the total duration of their LEV prescription: the short-term LEV group (LEV use < 30 days) and the long-term LEV group (LEV use ≥ 60 days). Patients prescribed LEV for 30-60 days were excluded from the final study population to clarify the impact of duration of LEV.
2. Assessed variables
We evaluated a comprehensive set of variables, including the duration of follow-up, sex, year of initial diagnosis, age at diagnosis, the interval between surgery and initiation of concurrent chemoradiation (CCRT), and time from diagnosis to surgery. Additionally, we evaluated the LEV administration before surgery, use of any AED before or after surgery, and presence of a seizure history before surgery, as indicated by ICD-10 codes G40 and R56. Other variables, such as the total number of days of AED prescription during the post-surgical follow-up, duration of hospitalization within the first 90 days post-surgery, initial dosage of LEV post-surgery, and number of adjuvant TMZ cycles received, were also considered.
3. Endpoints
The primary objective of our study was to assess the difference in OS between the long- and short-term LEV groups. Additionally, we explored whether LEV is associated with different survival outcomes, compared with other AEDs. We selected valproic acid (VPA) as a representative comparator among other AEDs.
4. Statistical analysis
Descriptive methods were utilized to provide an overview of patient characteristics and treatment approaches. Continuous variables were represented as means with standard deviations and categorized as counts and percentages. Differences between groups were determined using the independent t-test or Mann-Whitney U test for continuous variables, based on data distribution. We employed the chi-squared test or Fisher’s exact test, as appropriate, to analyze categorical data. We plotted Kaplan-Meier survival curves using the log-rank test to distinguish the OS rates between the groups. Both univariable and multivariable Cox proportional hazard models estimated hazard ratios with 95% confidence intervals (CIs) and considered potential confounders. All tests were performed in a two-sided manner, and statistical significance was set at p < 0.05.
Results
1. Population characteristics
We conducted a comprehensive nationwide population-based analysis of data from the HIRA database. First, 8,176 patients diagnosed with GBM (ICD-10 code: C71.x) between January 2007 and November 2021 were included. Among those patients, 5,311 underwent surgery and received CCRT within 6 weeks after surgery, conforming to the standard treatment protocols. After applying the exclusion criteria, 3,286 patients remained. Finally, 2,971 patients were included, after excluding those who received LEV for 30-60 days. Among this cohort, 1,319 were in the short-term LEV group (< 30 days), and 1,652 were in the long-term LEV group (≥ 60 days) (Fig. 1).
Table 1 presents the patient characteristics. The median follow-up time was 1.54 years (range, 0.98 to 2.93 years) for all patients. Men constituted 56.65% of the total population, with no significant difference in their proportion between the short-term (57.8%) and long-term (55.8%) LEV groups (p=0.270). The short-term LEV group had more patients diagnosed between 2007 and 2010 (47.1%), whereas the long-term LEV group had more patients diagnosed between 2015 and 2018 (59.2%, p < 0.001). The mean age was 56.25±12.70 years, with no significant difference between groups (p=0.530). The median time from surgery to CCRT was 25 days (interquartile range [IQR], 0 to 31) in both groups (p=0.598), whereas the median time from diagnosis to surgery was 2 days (IQR, 0 to 7) and 3 days (IQR, 0 to 10) in the short- and long-term LEV groups, respectively (p < 0.001).
LEV was administered before surgery in 8.49% of the short-term LEV group and 29.0% of the long-term LEV group (p < 0.001). A history of seizures before surgery was present in 21.5% and 32.3% of the patients in the short- and long-term LEV groups, respectively (p < 0.001). The median number of AED prescription days for follow-up after surgery differed significantly between groups for LEV, VPA, and other AEDs (all p < 0.001). The median length of hospital stay was similar between the groups. The first LEV dose after surgery varied significantly between the groups (p < 0.001). The mean number of adjuvant TMZ cycles was 4.77±1.69, with a slightly higher number in the long-term LEV group than in the short-term LEV group (p=0.010).
2. Survival analysis
Kaplan-Meier estimates showed a longer median survival in the long-term LEV group (19.06 months; 95% CI, 18.27 to 20.14) than in the short-term LEV group (17.64 months; 95% CI, 16.69 to 18.79; log-rank p=0.0046) (Table 2, Fig. 2A).
In the Cox proportional hazard models, univariable analysis indicated no significant difference in mortality risk between the long- and short-term LEV groups (unadjusted hazard ratio [HR], 0.81; 95% CI, 0.75 to 0.87; p < 0.001). After multivariable adjustment, the long-term LEV group was associated with a lower mortality risk, compared with the short-term LEV group (adjusted HR, 0.89; 95% CI, 0.82 to 0.96; p=0.002).
Other factors associated with mortality risk in the multivariable analysis were age and sex. A history of seizures before surgery was not significantly associated with mortality risk in the multivariable model. Length of stay and the time from diagnosis to surgery were not significantly associated with survival outcomes.
3. Subgroup analysis of patients without seizure history before surgery
Table 3 presents a subgroup analysis of OS in patients without a history of seizures before surgery. The short-term LEV group had a median survival of 17.61 months (95% CI, 16.59 to 18.79), whereas the long-term LEV group had a median survival of 19.71 months (95% CI, 18.46 to 21.03); the difference in survival between the two groups was significant (log-rank test, p=0.009) (Fig. 2B).
Cox proportional hazards analysis showed that a longer LEV prescription (≥ 60 days) was associated with a 13% reduction in the risk of mortality (adjusted HR, 0.87; 95% CI, 0.80 to 0.96; p=0.003). Higher age (adjusted HR, 1.02; 95% CI, 1.017 to 1.024; p < 0.001) and male sex (adjusted HR, 1.35; 95% CI, 1.23 to 1.48; p < 0.001) were associated with an increased risk of mortality. The longer length of hospital stay was also associated with a slightly increased risk of mortality (adjusted HR, 1.006; 95% CI, 1.004 to 1.007; p < 0.001). The time from diagnosis to surgery did not significantly influence the survival outcomes (adjusted HR, 0.999; 95% CI, 0.994 to 1.004; p=0.738).
4. Comparative influence of LEV and VPA on OS outcomes
Among the patients in the VPA group, the median survival was 16.53 months (95% CI, 15.31 to 18.53); patients in the long-term LEV group had a longer median survival of 19.06 months (95% CI, 18.27 to 20.14). The difference in survival between the two groups was significant (log-rank test, p=0.002) (Fig. 2C). In the Cox proportional hazard analyses, the univariable analysis indicated that the long-term LEV group had a reduced risk of mortality, compared with the VPA group (unadjusted HR, 0.83; 95% CI, 0.74 to 0.94; p=0.002). After multivariable adjustment, the long-term LEV group maintained a lower risk of mortality, compared with the VPA group (adjusted HR, 0.84; 95% CI, 0.74 to 0.95; p=0.004).
Other factors associated with increased mortality risk in the multivariable analysis included higher age (adjusted HR, 1.02; 95% CI, 1.017 to 1.025; p < 0.001) and male sex (adjusted HR, 1.26; 95% CI, 1.14 to 1.38; p < 0.001). The longer length of hospital stay was also associated with a slightly increased risk of mortality (adjusted HR, 1.005; 95% CI, 1.004 to 1.007; p < 0.001). A history of seizures before surgery and the time from diagnosis to surgery were not significantly associated with the survival outcomes in the multivariable analysis (Table 4).
Discussion
Our nationwide population-based study revealed that prolonged LEV administration (≥ 60 days) was associated with longer survival in patients with GBM receiving standard treatment. The long-term LEV group had a significantly longer median survival, compared with the short-term LEV group (19.06 vs. 17.64 months). Cox analysis showed that higher age and male sex were associated with a higher mortality risk, consistent with existing literature [18,19]. Long-term LEV usage remained a significant factor when these factors were adjusted for. The association between prolonged LEV use and longer survival was also observed in patients without a history of seizures and it was more pronounced in patients treated with LEV versus those treated with VPA. This suggests that the association between LEV and survival may not be solely attributable to its antiepileptic properties.
Several studies have investigated the survival benefits of LEV in patients with GBM. In 2015, a retrospective analysis of 103 patients with GBM revealed that 58 (56%) patients had been administered LEV for at least 3 months during their TMZ chemotherapy regimen [11]. Their cohort demonstrated significant increases in progression-free survival (PFS) by 2.7 months and in OS by 9 months. A subsequent 2020 study involved 322 patients with GBM diagnosed between 2004 and 2016, with all tumors surgically resected and pathologically confirmed as isocitrate dehydrogenase (IDH)-wild type [15]. The observed differences were 0.9 months in PFS and 3.6 months in OS. The prognostic factors identified included age, extent of resection, O6-methylguianine-DNA methyltransferase (MGMT) promoter methylation status, and Karnofsky performance status (KPS) score. The most recent study in 2022 analyzed a cohort of 460 patients with IDH-wild type GBM and evaluated the effect of LEV during standard chemoradiation [17]. Patients were categorized based on the duration of LEV treatment (full, partial, or none). A case-matched analysis of 54 patients per group indicated a significant survival advantage. Other prognostic factors, such as MGMT promoter methylation and extent of tumor resection, were also significant factors associated with survival.
However, not all studies reported positive outcomes. A comprehensive study including 1,263 patients with GBM aimed to determine the OS benefits of newer AEDs, including LEV, VPA, carbamazepine, oxcarbazepine, lamotrigine, and phenytoin [7]; no significant benefit was identified for any of the AEDs. Only 195 patients received LEV therapy, despite the large sample size. The study design was not planned to assess the effects of LEV on OS and lacked detailed data on LEV dosage and duration, inherently limiting the analysis of LEV efficacy in GBM treatment. In 2016, a study reported that neither VPA nor LEV provided survival benefits in patients newly diagnosed with GBM [18]. However, these studies have not sufficiently investigated the dose-dependent effects of LEV. In a survey of the members of the American Association of Neurological Surgeons and Congress of Neurological Surgeons in November 2015, only 13% of surgeons reported that the prescription of prophylactic postoperative AED was for longer than 6 weeks [19]. Therefore, it seems likely that patients included in previous studies may have been administered LEV for a short duration. Previous studies may have shown a survival benefit from LEV if many patients had been administered LEV for an extended duration.
Our study employed a large sample size to investigate this topic. We used the HIRA database, which encompasses the vast majority of data on patients with GBM in South Korea; therefore, our study can be considered a comprehensive, nationwide population-based study. Our study had a retrospective design and did not ascertain previously recognized prognostic markers, such as KPS, MGMT promoter methylation, and IDH mutation status; nevertheless, the likelihood of selection bias remains minimal owing to the unselected nature of our study population.
Several mechanisms have been proposed to explain the effect of LEV on the OS of patients with GBM. First, LEV may increase the sensitivity of GBM cells to TMZ. LEV is believed to increase the transcription of histone deacetylase 1, leading to the recruitment of the histone deacetylase 1/mSin3A complex. This affects the p53-binding site of the MGMT promoter, downregulating MGMT transcription and possibly enhancing TMZ’s antitumor activity [12-14,16]. Another theory focuses on chloride dynamics. Increased intracellular chloride concentrations have been linked with accelerated tumor progression. LEV’s effect on intracerebral gamma-aminobutyric acid (GABA) concentration and the GABA receptor is hypothesized to influence chloride equilibrium, subsequently aiding tumor progression control [20]. A third theory focuses on neuronal activity. The synaptic connection between presynaptic pyramidal neurons and postsynaptic glioma cells, which is mediated by glutamate, primarily through α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors, may promote glioma progression. LEV has anti-AMPA effects; therefore, its long-term use might inhibit neurogliomal synapses and suppress tumor progression [21-23]. This study could not determine the most likely mechanism or provide insights into a new explanation for the OS-extending effects of LEV; nevertheless, scientific evidence has been accumulated over time, considering previous experimental and clinical research outcomes. This research serves as supplementary proof for such hypotheses.
This study has some limitations. The study could not evaluate essential prognostic markers, such as the KPS score, MGMT promoter methylation, and IDH mutation status. Because of the inability to measure KPS, we used alternative indicators, such as the length of hospital stay. Although IDH mutation status was considered after 2016, histologically confirmed GBM cases were still included in the study. It is reasonable to assume that a similar proportion of IDH-mutant grade 4 gliomas were included before and after 2016, as IDH-mutant gliomas were only excluded from the GBM classification after 2021. As LEV was primarily used after 2010, other factors related to the passage of time may have contributed to the improvement in survival rates. However, given that no other treatment methods besides the standard TMZ therapy have been shown to improve survival, the results of this study suggest that long-term LEV use may be a significant factor contributing to the observed differences in survival duration.
We excluded patients who were prescribed LEV for 30-60 days because patients who received LEV for less than 30 days were considered to have been routinely maintained for a short period after surgery, whereas those who received LEV for more than 60 days were considered to have been administered LEV during the CCRT period. Patients who received LEV for 30-60 days were likely to have discontinued treatment due to side effects and were excluded to eliminate ambiguity. Notably, a significant difference was also observed when patients were divided into two groups based on a 60-day cutoff (data not shown).
As this is a retrospective study, the possibility of selection bias cannot be eliminated. To minimize immortal time bias, we only included patients who survived for at least 90 days after surgery. However, this method is not entirely foolproof, and caution should be exercised when interpreting the results.
Furthermore, we discussed the potential mechanisms underlying the effects of LEV; however, we did not validate these mechanisms, and our comparison predominantly focused on VPA, excluding other AEDs. The nuances of LEV dosage and duration remain underexplored. Further prospective studies are required to gain a comprehensive understanding of this phenomenon.
XML Download