Keywords
epilepsy - tumors - surgery
Tumor Classification
All types of primary and secondary brain tumor have the potential to cause seizures.
However, for the purposes of this review we shall discuss only the most common epileptogenic
tumors. Primary tumors can be divided into the rarer, benign-acting tumors, termed
long-term epilepsy-associated tumors (LEATs),[1] and the more common low-grade and high-grade astrocytic tumors encountered in neuro-oncological
practice.
Long-term epilepsy-associated tumors typically present in young patients with seizures.
Imaging shows cortically based, well-circumscribed lesions, often in the temporal
lobe, with no evidence of diffuse infiltration ([Fig. 1]). These tumors can be further subdivided into those of mixed glioneuronal origin
and those of glial origin. Glioneuronal tumors have the highest seizure rate, most
likely due to hyperexcitable regions of dysplastic neurons.
Fig. 1 (A) Coronal T2 fluid-attenuated inversion-recovery view of right-sided medial temporal
dysembryoplastic neuroepithelial tumor. (B) Axial T1-weighted magnetic resonance image with gadolinium contrast of right-sided
mesial temporal ganglioglioma. Note the small degree of gadolinium enhancement.
Low-grade gliomas are World Health Organization (WHO) grade II tumors, and comprise
diffuse astrocytoma (seizures seen in 50–81%), oligodendroglioma (seizures in 46–78%)[2] and mixed oligoastrocytoma. These slow-growing tumors infiltrate into surrounding
cortex and along white matter tracts, and ultimately undergo malignant transformation
into grades III and IV. The median survival for a diffuse astrocytoma is 8.2 years
([Fig. 2A]).[3] [Table 1] shows the epidemiological and pathological characteristics of the LEATs and low-grade
gliomas.[1]
[2]
[4]
Table 1
Table to show the epidemiology and pathophysiology of the long-term epilepsy-associated
tumors
|
Type
|
Origin
|
WHO Grade
|
Frequency
|
Most common location
|
Malignant transformation
|
|
Dysembryoplastic neuroepithelial tumor
|
Glioneural
|
I
|
15%
|
Temporal
|
Very rare[28]
|
|
Ganglioglioma
|
Glioneural
|
I
|
40%
|
Temporal
|
Rare
|
|
Gangliocytoma
|
Glioneural
|
I
|
Rare
|
Temporal
|
|
|
Angiocentric glioma
|
Glioneural
|
I
|
< 3%
|
Temporal
|
Unknown
|
|
Pilocytic astrocytoma
|
Glial
|
I
|
16%
|
Around third ventricle
|
Rare
|
|
Diffuse astrocytoma
|
Glial
|
II
|
16%
|
Frontal, temporal
|
Certain
|
|
Oligodendroglioma
|
Glial
|
II
|
11%
|
Frontal, temporal
|
Certain
|
|
Pleomorphic xanthoastrocytoma
|
Glial
|
II
|
2%
|
Temporal
|
Rare
|
Fig. 2 (A) Axial T2 fluid-attenuated inversion-recovery view of right-sided insular low-grade
glioma with extension into the temporal pole. (B) Coronal T1-weighted magnetic resonance image with gadolinium contrast of left-sided
temporal high-grade glioma. Note the rim-enhancing lesion causing local mass effect
and subfalcine herniation.
Common malignant tumors comprise the WHO grade III and IV glial tumors, and include
anaplastic astrocytoma, anaplastic oligodendroglioma, anaplastic oligoastrocytoma,
and glioblastoma multiforme (GBM). These tumors are considered to be less epileptogenic
than the LEATs and low-grade gliomas, with reported seizure rates of 22 to 62%,[2] although this may reflect the shorter life span of the patients (median survival
for patients with GBM with best treatment is 14.6 months.)[5] Like low-grade gliomas, these tumors also infiltrate along white matter tracts,
but they are far more aggressive, causing localized tissue destruction, ischemia,
and necrosis ([Fig. 2B]).
Pathophysiology
The pathophysiology of tumor epileptogenesis is multifactorial, and certain mechanisms
are probably more prevalent in certain tumor types. Long-term epilepsy-associated
tumors of glioneural origin are highly epileptogenic. This is thought to be a predominantly
direct effect, through hyperexcitable regions of dysplastic neurones developing inside
the tumor itself. However, sometimes there is an additional peritumoral component.
Upregulation of gap junctions locally may lead to peritumoral neural networks and
seizure propagation outside of the tumor.[2] Some cases have associated focal cortical dysplasia (FCD), recently classified by
the International League Against Epilepsy (ILAE) as FCD type IIIb.[6]
Gliomas have no intrinsic neural component, and so epileptogenesis must result from
tumor infiltration into surrounding cortex, with an indirect effect on the surrounding
neural networks. There is good evidence from magnetoencephalography, surface electroencephalography,
and stereo-electroencephalography that epileptiform activities in gliomas arise from
the peritumoral cortex.[7] There are many candidates for potential mediators of epileptogenesis, and they can
be broadly grouped into an epileptocentric and tumorcentric framework.[7]
Epileptocentric Causes
The epileptocentric hypothesis is that seizures arise from a change in the local excitability
of the peritumoral milieu. This is primarily due to a state of disequilibrium, caused
by the upregulation of excitatory glutaminergic transmission, coupled with decreased
GABA transmission. GABAergic transmission may play a more active role: Disturbances
in chloride homeostasis may lead to a pathological switch from an inhibitory hyperpolarizing
effect of chloride influx to an excitatory depolarizing effect of chloride efflux.
Other factors may also exacerbate local excitability, including pH shifts, increased
gap junction-mediated connectivity and blood–brain barrier disruption.[2] The epileptocentric framework for tumor epileptogenesis is most applicable to low-grade
gliomas, where there is slow growth, with gliosis and chronic inflammation in the
peritumoral cortex.
Tumorocentric Causes
The tumorocentric hypothesis is that epilepsy arises from the direct mechanical effects
of the space-occupying lesion. Local mass effect and edema lead to regional intracranial
hypertension and corresponding cerebral hypoperfusion. There is localized tissue destruction,
ischemia, and necrosis, and a simultaneous neoangiogenesis that may induce microhemorrhages
and further inflammatory changes. Structural reorganization and functional deafferentation
are well described, with the recruitment of nonmalignant cells, such as astrocytes
and microglia, further facilitating intercellular communication and seizure propagation.
The tumorocentric framework for tumor epileptogenesis is most applicable to high-grade
gliomas, which grow rapidly and cause significant local disruption.
Prognostic Factors
The natural course of epilepsy associated with brain tumors differs from other epilepsies;
surgical resection offers a potentially curative treatment. There are several clear
prognostic factors for seizure freedom following surgery. First, tumor type is important,
with good rates of seizure freedom seen in LEATs and WHO grade I gliomas. Surgery
on dysembryoplastic neuroepithelial tumors (DNETs) gives an average seizure freedom
of 88% at 5 years (range 58–90%); surgery on gangliogliomas gives seizure freedom
in 79% at 5 years (range 45–100%).[8]
Extended reviews of glioneural tumors indicate that the most important predictor of
seizure freedom in this group is total removal of the tumor.[9] Less important prognostic factors include reasonable preoperative seizure control,
absence of secondary generalized seizures, and a short history of epilepsy (<1 year).[9] Patient age and tumor location do not appear to be important. There is no evidence
that intraoperative electrocorticography (ECoG) makes any difference.[10]
Similarly, extended reviews of surgery for low-grade gliomas also indicate that gross
total resections are the most important prognostic factor in achieving seizure freedom.[10] In this context, gross total resection refers to the removal of the radiological
abnormality because the pathology is infiltrative and diffuse. As is the case with
LEATs, other predictors of seizure freedom include good preoperative control, absence
of generalization, and a short duration of epilepsy. With both LEATs and low-grade
gliomas in the temporal lobe, there is now reasonable evidence that adding an amygdalohippocampectomy
to a gross total resection further improves seizure outcome (87% vs. 79%).[10]
The prognosis for seizure freedom in patients with high-grade glioma is less good
because the pathology is destructive and widespread in the brain. The median survival
of this group is poor; therefore, there are no data to document long-term seizure
outcome following surgery.
Importance of Holistic Care
Importance of Holistic Care
The diagnosis of tumor-related epilepsy is often challenging for patients because
they carry with them the stigma of both epilepsy and tumor pathologies. Epilepsy diminishes
quality of life for several reasons, including the impact on driving eligibility,
employment and relationships, difficulty with independence, anxiety regarding seizures,
and reliance upon antiepileptic drugs and their associated side effects. The diagnosis
of a brain tumor brings additional concerns around the risk of malignant transformation
and long-term survival. These patients often have neuropsychological issues due in
part to the epilepsy and also to the use of antiepileptic drugs. Furthermore, changes
in seizure pattern can trigger concerns in patients regarding tumor transformation
or recurrence. Many would benefit from formal neuropsychological or neuropsychiatric
assessment, and this factor should always be considered when asking patients to weigh
up complex issues around their management. All patients with a diagnosis of tumor-related
epilepsy should be discussed in a multidisciplinary team environment, with input from
neurologists, neurophysiologists, neurosurgeons, neuroradiologists, and neuropyschologists.
The remainder of this review focuses on management of tumor-related epilepsy, covering
treatment aims, histological diagnosis, medical treatment, and surgical treatment.
Treatment Aims
The possible treatment aims in the management of tumor-related epilepsy are summarized
below:
-
Establish the histological diagnosis
-
Improve seizure control/achieve seizure freedom
-
Improve survival
-
Improve quality of life
The final aim, that is the improvement of quality of life, considers patients' posttreatment
lives, and takes into account any newly acquired neurologic or cognitive deficits
that may result from surgical treatment.
There is substantial overlap between these aims; for example, achieving seizure freedom
will likely lead to an improvement of quality of life, and can also be viewed as improving
survival, with a corresponding decrease in the risk of sudden unexpected death in
epilepsy (SUDEP). Similarly, there is increasing evidence that total resection of
low-grade gliomas can both improve seizure control and also prolong survival.[11]
The priorities of treatment will clearly vary from case to case, and will be informed
by patients' own ideas and concerns, as well as tumor type, pathophysiology, and likely
prognosis. As a rule, in patients with a low-grade glioma, oncological clearance takes
precedence, while for patients with LEATs it is the determination and removal of the
epileptogenic zone. Functional preservation is paramount in both groups ([Fig. 3]).
Fig. 3 Venn diagram to demonstrate the treatment aims and associated onco-epilepsy balance.
QoL, quality of life.
Importance of Histological Diagnosis
Importance of Histological Diagnosis
As previously mentioned, the histology and grade of the tumor is vital to determine
natural history and select the appropriate management strategy. Structural magnetic
resonance imaging (MRI) is probably sufficient to diagnose a LEAT, but is associated
with up to a 50% false-positive rate in predicting the WHO grade in astrocytomas.[12] To reduce the potential for misdiagnosis, there is therefore a clear indication
to achieve a histological diagnosis in patients with epilepsy and structural imaging
consistent with a glial tumor.
Aside from the risk of misdiagnosis, there is increasing interest in taking histological
samples for the purpose of molecular and genetic profiling. DNA methylation of the
O-methylguanine-DNA methyltransferase (MGMT) gene is well known as an independent
predictor of improved survival in patients with grade IV glioma who are receiving
temozolomide.[5] Similarly, the 1p19q deletion is associated with a more favorable survival in patients
with anaplastic oligoastrocytoma.[13] There is great value in identifying molecular markers in low-grade gliomas, to help
determine likely growth rate and response to treatment, with much interest in the
isocitrate dehydrogenase (IDH) mutational status.[14] In the future, the genetic profile will be added to the tumor histology and grade
to help to decide prognostic and treatment paradigms.
Medical Treatment Strategies
Medical Treatment Strategies
Medical treatment of epilepsy is required for surgical candidates to control seizures
until surgical efficacy has been established. Drug treatment is also required in patients
with tumor-related epilepsy, who are not candidates for surgical treatment. Treatment
is often challenging, with up to 24% of patients requiring three or more antiepileptic
drugs.[15]
There are few randomized controlled trials comparing the efficacy of antiepileptic
drugs in the treatment of this patient group. The choice of medication is therefore
mainly determined by side-effect profile and drug interactions. As a general rule,
antiepileptic drugs that induce cytochrome P450 hepatic enzymes should be avoided
in patients who are likely to undergo chemotherapy (i.e., patients with high-grade
glioma), as they can enhance the metabolism of chemotherapeutic agents, and they are
often a contraindication to participating in clinical trials.
Interestingly, there is increasing evidence that some antiepileptic drugs may have
direct antineoplastic actions that can improve survival in patients with high-grade
glioma. Retrospective studies have shown that the use of valproate in patients with
glioblastoma multiforme is associated with better survival, and that this effect depends
upon treatment with temozolomide.[16] Similarly, in vitro studies have shown that levetiracetam can suppress MGMT expression,
with potential implications for the augmentation of temozolomide in adjuvant treatment
of glioblastoma.[17]
The current first-line drugs in neuro-oncological practice are valproate and levetiracetam;
however, many other antiepileptic drugs are also well tolerated, with few meaningful
drug interactions, supporting their use as an adjunctive treatment in patients with
tumor-related epilepsy.
Surgical Treatment Strategies
Surgical Treatment Strategies
Long-term epilepsy-associated tumors rarely undergo malignant transformation; low-grade
gliomas inevitably progress to higher grades over the course of the disease; high-grade
gliomas are already malignant. In each case the surgical strategy should have clear
aims, informed by the concept of an onco-epilepsy balance tailored to that individual
patient (see [Fig. 3]). In this section, we describe how best to approach these three groups of patients,
striking the right onco-epilepsy balance. In each case, we assume a basic presurgical
evaluation has taken place, and that seizure semiology, video telemetry, and neuropsychology
are broadly concordant with the lesion in terms of likely seizure onset.
Long-Term Epilepsy-Associated Tumors
Although there are isolated reports in the literature of malignant transformation,
LEATs generally behave in a very benign fashion.[18]
[19] Because these patients are likely to lead long lives, the priorities of surgical
treatment should be to improve seizure control, but not at the expense of creating
a neurologic deficit. The onco-epilepsy balance in this group should therefore be
heavily biased in favor of curing the epilepsy. Here we deal with solitary LEATs first,
followed by a discussion on LEATs in the context of dual pathology. [Fig. 4] gives an overview of the surgical treatment of LEATs.
Fig. 4 An overview on the surgical treatment of long-term epilepsy associated tumors. AmHipp,
amygdalohippocampectomy; EEG, electroencephalogram; FCD, focal cortical dysplasia;
IC, intracranial; MTS, mesial temporal sclerosis.
Solitary Long-Term Epilepsy-Associated Tumors
Long-term epilepsy-associated tumors are often small, superficial, and well-circumscribed
lesions, making them particularly amenable to total surgical resection. In theory,
the pathophysiology of these tumors means that the epileptogenic zone should arise
from within the tumor, and excellent results can be obtained in terms of seizure freedom
following total lesionectomy. Interestingly, there is now good evidence that “solitary”
glioneural tumors may in fact be associated with cryptogenic focal cortical dysplasia.[20] Despite this, resections informed by invasive EEG (IC-EEG) do not seem to improve
seizure freedom outcomes in this group. In most cases, the most pertinent question
is therefore whether the lesion extends into eloquent cortex, and whether a total
resection is feasible.
If the LEAT is small, solitary, and resides in noneloquent cortex, it is reasonable
to proceed with a total lesionectomy. This can be done as an awake craniotomy with
intraoperative functional mapping if there are any concerns regarding functional boundaries.
There is no need in these cases to perform preresection IC-EEG to map out the likely
epileptogenic zone because it likely resides within the tumor. There are considerable
risks associated with IC-EEG, and there is no added benefit if the likely outcome
will be a total lesionectomy. However, in the rare instances when the patient continues
to have seizures following surgery, IC-EEG can then help to identify the epileptogenic
zone in the peritumoral cortex. Further resection may ultimately demonstrate evidence
of MRI-negative pathology, such as small residual tumor or focal cortical dysplasia.
If the LEAT resides or encroaches on eloquent cortex, the benefits of a total lesionectomy
need to be balanced against the lifelong risk of neurologic deficit. In these cases
we recommend pausing to consider whether surgical management is appropriate at all.
If the patient wishes to continue along the surgical route, the next step is IC-EEG
to map out the seizure-onset zone and irritative zone, and also to map the functional
zone.[21] Before implantation, these patients should undergo a thorough noninvasive presurgical
evaluation, with data from advanced multimodality imaging used to inform the invasive
monitoring.[22] The aim of IC-EEG is to localize a part of the lesion where the seizures begin,
and that can be safely resected, that is to inform a targeted subtotal lesionectomy.
The relative strengths and weaknesses of subdural grid coverage versus stereo-EEG
in general are discussed elsewhere.[23] In these cases, tumors are often superficial, and mapping of the functional zone
is limited to the surrounding cortex. We therefore recommend subdural grids for IC-EEG,
because they provide better spatial resolution and make subsequent resection more
straightforward to plan. In patients who progress to lesionectomy, there are clear
advantages with awake craniotomy to confirm the extraoperative functional mapping.
Long-Term Epilepsy Associated Tumors and Dual Pathology
Mesial Temporal Sclerosis
Although LEATS can occur anywhere in the supratentorial compartment, they are most
commonly found in the temporal lobe. There is a well-established link with mesial
temporal LEATs and mesial temporal sclerosis, which is often not visible on the MRI.
This presents a particular dilemma in surgical management; total lesionectomy coupled
with removal of the mesial temporal structures (amygdala and hippocampus) gives superior
results for seizure freedom over just lesionectomy,[9] but is associated with higher risks of postoperative neuropsychological deficits
in visual and verbal memory.
As stated previously, each patient should undergo rigorous presurgical evaluation,
including neuropsychological testing to predict physiological reserve and the likely
impact on quality of life, and surgical management should be tailored to individual
needs. In general, if the lesion is on the nondominant side, the cognitive risk associated
with an amygdalohippocampectomy is considerably less, favoring the removal of both
the lesion and the mesial temporal structures, with high chances of achieving seizure
freedom. However, if the lesion is on the dominant side and the patient has high physiological
reserve, the cognitive risk associated with amygdalohippocampectomy is greater, and
can significantly affect the patients' quality of life. In these instances, the pragmatic
approach is a total lesionectomy, preserving the mesial structures if they appear
structurally and functionally normal. It should be discussed preoperatively with these
patients that this approach carries a slightly lower chance of achieving seizure freedom.
However if the patients are not seizure-free following surgery, further surgical options
remain open. This may be extension of the resection to include the mesial structures
if implicated, or IC-EEG to determine the epileptogenic zone in surrounding cortex.
Focal Cortical Dysplasia
The ILAE classification of focal cortical dysplasia includes type IIIb, where cortical
lamination abnormalities lie concurrent to a tumor.[24] Most commonly this occurs with LEATs, although they can also occur with low-grade
gliomas.
The treatment challenge in these cases is to determine whether the epileptogenic zone
arises from the tumor or the peri-tumoral dysplasia. For small tumors in non-eloquent
areas with limited focal cortical dysplasia, the precise identification of the epileptogenic
zone is probably unnecessary, and the pragmatic approach is to perform a lesionectomy
and extended resection of the peri-tumoral cortex to include the adjacent dysplasia.
For larger tumors with encroachment into eloquent cortex, this extended resection
is not possible and precise determination of the epileptogenic zone is crucial. In
these instances, IC-EEG is the first step to determine likely epileptogenic and functional
zones. Further surgical treatment can then be tailored to that individual according
to the findings of the invasive study.
Low-Grade Glioma
Malignant transformation in low-grade glioma is certain, with a median survival of
8.2 years.[3] Although the prognosis of these tumors remains poor, there have been some advances
in our understanding that inform modern surgical management.
Historically, it was thought that the natural history of these tumors was poor irrespective
of the surgical treatment. However, it is becoming increasingly accepted that the
extent of resection is a major predictor of length of survival, with successively
greater improvements associated with extent of resection over 70%.[11] A further change to our understanding is in what constitutes a total resection.
The convention has been to describe this radiologically as the volume defined by gadolinium
enhancement on the T1-weighted MRI. However, it is well established that tumor cells
infiltrate white matter beyond these margins, as evidenced by the large signal change
on T2 fluid-attenuated inversion recovery (FLAIR) MRI. The onus in these patients
is therefore shifting away from biopsy toward total resection, and even supratotal
resection, where the resection is only limited by functional boundaries.[25] Thus, the onco-epilepsy balance in this group has shifted toward maximal oncological
clearance, with the additional benefit of large-volume histological samples for accurate
grading and genetic profiling.
Total and supratotal resections depend on the identification and preservation of functional
boundaries. This can be achieved in two ways. One way is to use IC-EEG to perform
extraoperative mapping of eloquent cortex. Subdural grids are best suited to this
because they generate good spatial coverage, and their placement can be informed by
functional MRI and gyral anatomy on 3-dimensional computer models.[22] The advantage of this technique is that the mapping can be done under controlled
conditions, with the patient fully conscious. This is often employed in the determination
of the functional zone in children ([Fig. 5A]).
Fig. 5 (A) Intraoperative photograph of subdural grid over lesion, a dysembryoplastic neuroepithelial
tumor (DNET), for extraoperative mapping; top left: photograph of left frontoparietal cortex with DNET straddling motor cortex, top right: three-dimensional (3D) computer model of cortex (pink), lesion (red), cortical veins (cyan), hand transcranial magnetic stimulation (green), and leg transcranial magnetic stimulation (orange); bottom left: intraoperative photograph of left frontoparietal cortex with overlying subdural
grid; bottom right: 3D computer model of cortex, lesion, implanted electrode contacts (yellow). (B) Photograph of intraoperative mapping by cortical stimulation in an awake patient.
However, there are several disadvantages to this technique. Usually, subdural grids
have 1-cm spacing between contacts, making for limited accuracy. The implantation
of a subdural grid carries substantial risks of infection and hemorrhage, and commits
the patient to undergoing two surgical procedures. Perhaps most importantly, extraoperative
mapping only samples cortex and provides no information on the underlying white matter
tracts. This is crucial; White matter connectivity is the route that the pathology
migrates along, and also represents the limitation of brain plasticity potential.
The alternative mapping method is intraoperative mapping, eliciting responses in the
awake patient during surgery using cortical and subcortical stimulation. The key advantage
with this technique is that it allows for the functional mapping of the underlying
white matter tracts. A recent meta-analysis, based on 20 years of glioma surgery and
over 8,000 cases, has shown that the use of intraoperative mapping significantly increases
the extent of resection and also reduces the rate of postoperative neurologic deficits
in eloquent areas.[26] It is possible to map complex functions such as language, spatial cognition, calculation,
judgment, and executive functions with this technique. This requires an integrated
approach, with specialist testing, the correct organizational framework, and excellent
anesthesia ([Fig. 5B]).[27]
High-Grade Glioma
In patients with high-grade glioma, seizure control is rarely the key issue. Treatment
aims for these patients is to prolong survival and maintain quality of life. The optimum
treatment is surgical resection followed by adjuvant chemoradiotherapy. However, seizures
are obviously associated with a poor quality of life, and the potential to improve
seizure control should be factored into the decision to perform a radical tumor resection.
Summary
In summary, it is simplistic to assume that tumor-related epilepsy should always be
managed by lesionectomy. The treatment depends on tumor type and associated pathophysiology,
and should be tailored to individual patients, based on discussions around feasible
aims. The onco-epilepsy balance distinguishes LEATs from low-grade gliomas. The emphasis
with LEATs is on epilepsy control; this group has an excellent prognosis, with the
onus on early diagnosis and surgical management to achieve improved rates of seizure
freedom. The treatment of LEATs is complicated by the possibility of dual pathology.
By contrast, the emphasis with low-grade gliomas is on oncological clearance; this
group has a worse prognosis, but also relies on early diagnosis coupled with supratotal
resection to achieve improved survival rates. In both groups, functional preservation
is critical. The use of IC-EEG to inform resection should be rationed to particular
cases, where there is a clear practical benefit that will inform any subsequent resection.
Medical management of tumor-related epilepsy is an important adjunct to surgical management,
with the choice of antiepileptic drugs largely determined by side-effect profile and
drug interactions. Care of these patients is complex, and requires a multidisciplinary
team approach.
