Introduction

Brain tumors are serious medical concerns that result in significant morbidity and mortality, as well as high utilization of healthcare systems [1]. There are many different types of brain tumors, including gliomas, astrocytomas, meningiomas, and metastases, among others [2]. Some of the most serious and devastating complications of brain tumors are seizures. Seizures occur in approximately 20–45 % of brain tumor patients resulting in significant morbidity and reduction in quality of life [3]. Therefore, effective seizure prophylaxis is required to ensure the best possible patient outcomes are achieved.

One of the most important factors influencing the use of antiepileptic drugs (AEDs) in these patients is craniotomy for tumor reduction or removal. The incidence of seizures is estimated to be 15–20 % for patients undergoing non-traumatic, supratentorial craniotomy [4]. Other factors that influence development of seizures and choices of agent include extent of tumor resection in craniotomy (complete or partial), previous history of seizures, tumor location, or rate of tumor growth [5].

The limited evidence available precludes development of clinical practice guidelines or consensus statements regarding the use and choice of AEDs for brain tumor patients. However, current clinical trends show newer AEDs becoming first-line options [5]. This is likely a result of greater adverse effect and drug interaction potential from other agents, such as phenytoin [6]. Levetiracetam (LEV) is one of these new AEDs and believed to modulate synaptic neurotransmitter release by binding synaptic vesicle protein SV2A in the brain [7]. It has advantages over older agents, such as low hepatic metabolism and better tolerability in terms of adverse effects. These properties result in decreased potential for drug interactions [8]. As such, these considerations make LEV a favourable AED for use among clinicians worldwide.

To date, there is no high quality systematic review or comprehensive summary of evidence assessing the role of LEV for brain tumor patients. One systematic review applied strict inclusion criteria and only included one study [9]. Additionally, a second identified more studies but did not address study quality or critically analyze results [10]. As LEV is rapidly becoming clinicians’ first choice as an AED for these patients, a comprehensive and critical review is required to guide decision-making and further research aims. Therefore, our objective was to identify, summarize, and evaluate the literature pertaining to the efficacy and safety of LEV for preventing seizures in patients with brain tumors with or without planned craniotomy.

Methods

A literature search was completed using the databases PubMed (1948 to December 2015), EMBASE (1980 to December 2015), Cochrane Database of Systematic Reviews, and Google Scholar. The search terms employed for electronic database searching were “levetiracetam” OR “anticonvulsant” OR “keppra” OR “antiepileptic drug” combined with “brain tumor” OR “brain neoplasm,” using AND to combine search term categories. The search was limited to studies in humans and those published in English. Manual searching of the reference lists of identified articles and review articles were also used to capture any records not accounted for in the electronic search. The search was completed by one investigator and repeated by a second.

Studies were included in the systematic review based on predefined inclusion criteria: randomized controlled trials (RTCs) or observational studies (prospective or retrospective) that reported seizure frequency data of LEV either as monotherapy or combination therapy with other agents in patients presenting with tumors or metastases in the brain. Studies were excluded if seizure frequency could not be extracted for LEV users, or if data reported was solely from case reports. Two investigators assessed each identified study for inclusion and resolved any discrepancies through discussion.

Data extracted included study design, population, interventions or procedures, outcomes, and findings related to the primary outcome of seizure frequency. Data was extracted by one investigator and verified by a second investigator. Data were also extracted for risk of bias assessments according to the Cochrane Collaboration’s risk of bias assessment tool for RTCs [11]. Any study perceived to be at high risk of bias in any category was deemed to have an overall high risk of bias. Observational studies were assessed using the same tool, while accounting for design-specific biases within the ‘other biases’ category. Two investigators completed this independently. Any discrepancies were resolved through discussion.

Results

The literature search produced 2300 electronic hits and 12 hits from manual searching, as shown in Fig. 1. After title and abstract review, the full text versions of 44 articles were downloaded for review. After assessment against inclusion criteria, a total of 21 articles were included in the systematic review. Reasons for article exclusions are given in Fig. 1. The final included studies consisted of 3 RTCs [12–14], seven prospective observational studies [15–21], and 11 retrospective observational studies [22–32]. Characteristics and results from each included study are given in Tables 1 and 2.

Fig. 1
figure 1

Flow diagram for study selection and inclusion

Full size image
Table 1 Study characteristics and results from randomized controlled trials
Full size table
Table 2 Study characteristics and results from observational trials
Full size table

Risk of bias assessments are given in Table 3. All studies were deemed to be at high risk of bias. Two of three RCTs were deemed to have adequate sequence generation, however all RCTs did not report adequate allocation concealment or blinding. Two of three RCTs reported complete outcome data. Other biases were largely unclear. The observational studies all scored as being at a high risk of bias. Prospective studies generally were at risk of attrition bias from incomplete outcome data. Both prospective and retrospective studies had an unclear risk of other biases, as confounding factors were not always evident or accounted for. A major point to consider is that observational studies may be confounded by interventions such as tumor reduction or removal and the typical decline in seizure activity over time after craniotomy.

Table 3 Risk of bias assessments for identified studies
Full size table

Data from randomized controlled trials

Three RTCs were identified that assessed efficacy and safety of LEV in the target population. In 2009, Lim et al. reported results from a phase II pilot study assessing the feasibility of switching patients from phenytoin (PHT) to LEV monotherapy following craniotomy for glioma-related seizure control [12]. A total of 29 patients were randomized in a 2:1 fashion (LEV:PHT) and followed for 6 months. Upon follow up, data was only available for 15/20 patients in the LEV group and 8/9 in the PHT group. Of these patients, 13/15 (87 %) receiving LEV and 6/8 (75 %) receiving PHT remained seizure free at 6 months. Adverse effects were similar between groups, with coordination difficulties reported more commonly in the PHT group.

A second study evaluated LEV versus PHT during and after craniotomy for brain tumors (glioma, metastasis, meningioma, others) [13]. A total of 147 patients were randomized to receive LEV (n = 74) or PHT (n = 73) until postoperative day 7. The primary outcome was occurrence of seizures. No seizures occurred in any patient during surgery. Twelve patients developed seizures after surgery (1.4 % of those taking LEV vs. 15.1 % of those taking PHT, p = 0.005). However, due to the small sample size and seizure frequency, the odds ratio of postoperative seizures with LEV versus PHT was difficult to interpret [OR 12.77, 95 % CI 2.39–236.71]. Therapy was withdrawn due to adverse events in 5 (6.8 %) of PHT patients yet none of the LEV patients.

The third RCT evaluated LEV against pregabalin monotherapy in patients with gliomas [14]. Patients were eligible if they had at least one previous seizure and were randomized to receive either LEV (n = 25) or pregabalin (n = 27). The primary endpoint was survival free of a composite endpoint that included status epilepticus, 2 seizures with impaired consciousness, need of a second agent, or need to discontinue study drug. At 1 year follow up, 9/25 (36 %) of those taking LEV and 12/27 (44 %) of those taking pregabalin failed therapy. The composite endpoint was driven by the need to interrupt study drug (7/9 in LEV group and 7/12 in pregabalin group) and this was mostly due to adverse effects.

Data from prospective observational studies

Seven prospective observational studies were identified that reported seizure frequency as an outcome with LEV therapy. Six studies included comparisons between baseline and follow up seizure frequencies. Six studies reported data for up to 30 patients, while one study enrolled 176 patients (Table 2).

Studies reporting outcomes from prospective observational studies before and/or after surgical procedures

Bahr et al. enrolled 30 brain tumor patients with at least one seizure and planned neurosurgery (resection or biopsy) in a single-arm prospective study [15]. Patients received oral LEV up to 4 weeks before surgery and then IV LEV (with step down to oral therapy when indicated) for up to 4 weeks after surgery. A total of 25 patients were fully evaluable with 100 % having no seizures during pre-surgical phase, 88 % in the 48 h period post-surgery phase, and 84 % between 48 h post-surgery up to 4 weeks. No major safety concerns were noted.

Usery et al. enrolled 17 patients with operable brain tumors and a history of at least one witnessed seizure [16]. Patients were continued, converted to, or started on LEV monotherapy within 6 h following surgery for up to a minimum of 48 h. Patients had an average of 3.5 (range 1–22) seizures preoperatively. Post-operatively 16/17 (94.1 %) of patients had complete seizure control while in hospital and 11/12 (91.7 %) post-discharge. Five patients were lost to follow up and no outcome data post-discharge was available.

Rosati et al. evaluated seizure outcomes in 176 patients with glioma presenting to a neurosurgery clinic for follow up procedures [17]. Of the 176 patients, 82 had been diagnosed as having epilepsy with durations ranging from 13 months to 4.2 years. After mean follow up of 13.1 months (range 10 months to 2.9 years), 75 of 82 patients (91 %) were seizure free. Seventy-three of these patients were taking LEV. In patients experiencing recurrent seizures, dosage increases of LEV were given to achieve seizure relief. Only transient somnolence was documented as an adverse effect in four patients.

Studies reporting outcomes from prospective observational studies without clear surgical intervention for all patients

Maschio et al. completed a prospective case series of 29 patients with brain tumors referred to a specialized tumor-related epilepsy center with at least two seizures per month [18]. All patients were initiated or converted to LEV monotherapy and followed for 12 months of follow up. At 12 months, 15 evaluable patients remained with 1 having ≥50 % reduction in seizure frequency and 14 (93.3 %) remaining seizure free. An intention-to-treat population (n = 29, mean follow-up of 8.6 months) analysis found 21 patients seizure free (72.4 %), 7 (24.1 %) patients with ≥50 % reduction in seizure frequency and 1 (3.5 %) with stable seizure frequency. One patient developed a side effect (restlessness) that warranted discontinuation of LEV therapy.

Maschio et al. reported results from a prospective cohort assessing 19 patients with brain tumors and seizures [19]. At baseline, patients were experiencing seizures at daily to monthly frequencies. LEV was added to AED regimens in all patients and median follow up occurred over 20 months (range 7–50). At the end of follow up, 9 (47.4 %) of patients were seizure free (seizure free period range 7–33 months), 5 (25 %) reported improvement from daily to weekly. Seizure frequency did not change in 4 (21 %) of the patients and increased in 1 patient. No adverse effects related to LEV were noted.

Wagner et al. reported results from a study assessing the efficacy of LEV in 26 patients with primary brain tumors who had persisting seizures, adverse effects from other AEDs, and/or potential drug interactions with chemotherapy regimens [20]. LEV was added as combination therapy in 25 patients. For 20 patients with persisting seizures, ≥50 % reduction in seizure frequency was achieved in 13 (65 %) of patients over a mean follow up period of 11.8 months. Four of these patients became completely seizure free. The remaining 6 patients using LEV for indications of adverse effects from other agents or drug interactions became seizure free at the end of follow up. Adverse effects occurred in 9 (35 %) of the patients and most frequently were fatigue, somnolence, and dizziness.

Studies reporting outcomes from prospective observations studies assessing patients with brain metastases

Maschio et al. completed a prospective observational study of 48 patients with seizures related to brain metastases but only 30 returned to study site and had outcome data available [21]. At first visit, patients received LEV (n = 6), oxcarbazepine (n = 16), or topiramate (n = 8) according to usual practices and then followed until death (mean duration of follow up was 6.1 months). Baseline mean seizure frequency in the LEV group was 12.2 (±21.73 SD) and was reduced to 5.33 (±12.1 SD) at the last visit preceding patient’s death (p = 0.027). The oxcarbazepine and topiramate groups also had significant reductions in seizure frequencies. No severe adverse effects were noted. In the LEV group, 1 patient experienced rash and 1 patient experienced restlessness.

Data from retrospective observational studies

Eleven retrospective studies were identified that reported seizure frequency associated with LEV therapy [22–32]. Study data and results are given in Table 2.

Data from studies assessing LEV use in patients undergoing surgical tumor resection or biopsy was generally in favor of LEV as a first-line agent. Seven studies directly assessed surgical outcomes [22–28]. Two studies found no differences between LEV and PHT in terms of seizure frequency post-surgery [25, 27]. One study found similar post-surgical seizure rates between LEV and valproic acid [23], while another reported the combination of these two agents to be highly effective (although no statistical comparisons were completed) [24]. In high-risk patients undergoing surgery (pre-existing seizures, supratentorial meningioma, supratentorial low grade glioma), one study reported a 7-day post surgical seizure rate of 7.3 % associated with LEV use [28]. Finally, two studies assessed LEV use in patients with no planned surgical interventions and found statistically significant reductions in seizure frequency after LEV initiation [29, 30].

Reporting of safety outcomes was variable. One study showed significant reductions in adverse effect rates with LEV when compared to valproic acid [23]. Additionally, another study reported significant decreases in adverse effects requiring discontinuation of therapy with LEV versus PHT [27]. Finally, two studies reported the most common adverse effect with LEV to be somnolence, which was deemed to be mild and occurred in 23–37 % of patients [29, 30].

Discussion

This systematic review identified 21 studies that reported seizure frequency outcomes associated with LEV use in patients with brain tumors. The data obtained from RCTs is encouraging but must be interpreted cautiously. However, this is the best evidence available to date and interpretations can guide clinical decision-making. The study by Iuchi et al. provides strong evidence that LEV is likely effective and safe up to 7 days post-craniotomy, as compared to PHT [13]. The study was limited by sample size and outcome frequency (as demonstrated by the very wide confidence interval) and so conclusions regarding better efficacy compared to PHT are only speculative. Subsequent studies should include a longer follow up duration (6 months to 1 year) to better assess the long-term efficacy and safety of LEV. The two other smaller RCTs demonstrated that LEV was no worse than study comparators, although data should be considered preliminary only [12, 14]. Outcome rates across all RCTs differed greatly, suggesting high heterogeneity between study settings, populations, and designs.

While observational studies are prone to bias and confounding, both the prospective and retrospective studies reported efficacy outcomes in favour of LEV. Although positive, these finding must be interpreted cautiously as it is possible that confounders such as tumor reduction or removal and decreases in seizure activity over time may have influenced results in favour of LEV. Therefore, long-term benefits of LEV in this population are still unknown.

No major safety concerns were noted across all studies. The most common adverse effect noted was somnolence and was typically mild in nature. Discontinuations due to LEV adverse effects were uncommon, especially when compared to studies reporting the same outcome with PHT and valproic acid [12, 13, 23, 27]. These findings are not surprising, as it is well known that LEV is better tolerated than other, older agents [8].

Practice implications of our findings are relevant to the use of newer AEDs such as LEV for seizure prophylaxis in brain tumor patients, as compared to older, traditional agents. Consensus statements and clinical practice guidelines should consider the evidence presented in this review to direct future decision-making. Specifically, LEV was found to be a suitable therapeutic alternative in terms of efficacy, as compared to other agents. Additionally, enhanced tolerability and lack of drug interactions and need for therapeutic drug monitoring support its use as a valid option for these patients.

The major limitation of this review was the poor quality of identified studies. Every article was found to be at a high risk of bias. This was not surprising, due to the challenges designing studies in this population in terms of obtaining enough patients for adequate power and highly individualized nature of treatments and medication responses. Future studies can limit bias by adhering to good randomization, allocation concealment, and blinding principles. Additionally, prospective observational studies can make greater attempts to avoid attrition bias.

Conclusions

Efficacy data reviewed for LEV supports its use as a first line agent for patients with brain tumors. No worsening efficacy was noted against any other agent and seizure frequencies were commonly reduced with its use. This conclusion is supported by clear benefits in safety outcomes with LEV versus other agents. A future well-designed RCT or prospective observational study is warranted to further evaluate the role of LEV and other AEDs in brain tumor patients despite difficulties and limitations proposed by the disease/population. Finally, clinicians caring for patients with brain tumors should use the evidence presented in this review, along with strict patient monitoring and reassessment, to optimize seizure prophylaxis therapy and achieve the best possible patient outcomes.