The syndrome of frontal lobe epilepsy

Frontal lobe epilepsy (FLE), estimated by some to affect perhaps 600,000 individuals in the United States alone, ^[1] poses a number of unique diagnostic and management problems ^[2-7]. Frontal lobe seizures produce a variety of behaviors, some with bizarre features that may suggest a diagnosis of pseudoseizures ^[8]. EEG can be normal in either the interictal state or during seizures, ^[2,3] and neuroimaging studies are often unrevealing. Moreover, the rapidity with which frontal lobe seizures spread ^[4,9] and the propensity for bilateral representation of interictal discharges ^[10] often hinder the accurate localization of frontal lobe epileptogenic lesions.

Although methods for diagnosis and surgical management of temporal lobe epilepsy have improved tremendously in recent years, ^[11] refractory FLE remains less well defined. Most papers have examined relatively few patients with proven FLE, and there is a paucity of data in surgically treated patients. The purpose of this paper is to review the clinical characteristics, diagnostic evaluations, and treatment results in a series of patients treated surgically for refractory FLE. We hoped to gain further insight into FLE, to better define it so that it can more easily be distinguished from other forms of epilepsy, and to examine what features might offer the greatest chance of surgical success.

Methods. Of 289 patients surgically treated for intractable epilepsy at the Graduate Hospital, Philadelphia, PA, between 1986 and 1993, 16 were diagnosed as having FLE and included in this review. Presurgical evaluation included video-EEG monitoring, neuropsychologic testing, brain MRI, intracarotid amobarbital test, and invasive EEG when indicated ^[12]. Scalp EEGs were continuously recorded with sphenoidal and 10-20 System electrodes with 16 to 32 channels per patient with on-line spike and seizure detection (Telefactor Corporation, Conshohocken, PA). Behavior was recorded continuously on video cameras to supplement observations by nurses.

Patients who had frontal lobe structural lesions (eg, tumor) on MRI all had resective surgery without prior intracranial EEG monitoring. In all but one case, preoperative scalp/sphenoidal monitoring was performed. Resection strategy aimed at removing all the lesion and a margin of adjacent cortex defined by gross appearance and electrocorticography.

Patients whose MRI did not show frontal lobe structural lesions all had chronic scalp/sphenoidal EEG monitoring followed by intracranial EEG before resection. Intracranial EEG was performed with a combination of frontal lobe subdural strips and grids, temporal lobe depth electrodes and subdural strips, and parietal and occipital subdural strips or grids, placed in accordance with clinical symptoms and scalp EEGs. We placed intracranial electrodes bilaterally, to sample mesial, dorsolateral, frontopolar, and orbitofrontal cortex. Between 32 and 64 channels of EEG were recorded. The area resected included regions with both ictal and interictal abnormalities.

Operative resection was performed by a subpial technique, tailored to the interictal and ictal EEG abnormalities. Electrocorticography helped define the posterior margin of the resection in anterior frontal lobectomy or defined the perimeter of resection in a topectomy. In some patients, subdural electrode strips were placed over adjacent cortex at the close of an initial resection for additional video-EEG monitoring if a staged excision was considered. Intraoperative or extraoperative electrical brain stimulations were used when appropriate.

Anterior corpus callosotomy was performed as the sole procedure if a unilateral frontal lesion or seizure onset could not be defined. In some patients, an anterior callosal section accompanied the frontal resection; this procedure was performed when technically feasible as part of an effort to improve outcome beginning in the late 1980s.

Four seizure outcome classes were defined: class 1 = completely seizure-free since surgery or some postoperative seizures, but now seizure-free for greater than 1 year or auras only; class 2 = less than three seizures per year or nocturnal seizures only; class 3 = meaningful reduction in seizures, defined as greater than or equal to 80% decrease in seizure frequency; class 4 = less than 80% seizure reduction in seizures or no improvement.

Outcome data were updated by regular (at least yearly) office visits or telephone contact with patients and their referring physicians. All outcome data in this report indicate current outcome class. All patients in this study have continued to take anticonvulsant medications postoperatively, although reductions in dosage or number of medications have been prescribed.

Results. Operative procedures. Of the 16 patients, 14 had frontal resections, seven in conjunction with an anterior corpus callosotomy. The two remaining patients had anterior corpus callosotomy without cortical resection; in one patient, bilateral independent frontal epileptogenic lesions were found, and in the other, seizures could not be lateralized. Details are provided for each patient in table 1. Of the 14 patients who had cortical excisions, (1) three patients had dorsolateral frontal excisions, one in conjunction with an orbitofrontal resection and another with an orbitofrontal resection and posterior frontal subpial cortical isolation; (2) six patients had frontal pole resection, two in conjunction with mesial resections and one with a subsequent frontal lobectomy sparing motor cortex; (3) three patients underwent mesial frontal resections, two in conjunction with orbitofrontal resections; and (4) two patients underwent primary orbitofrontal resections, one in conjunction with an ipsilateral amygdalotomy.

Historical features. Patients varied widely in their historical features (table 1). Seizures began at an average age of 13.3 years, with a range from 3 to 38 years. Most patients had some identifiable risk factor, including tumor or vascular malformation (9), trauma (1), or encephalitis (1). Five patients had no risk factors for epilepsy. Presence or absence of a risk factor did not correlate with age of seizure onset. Before surgery, seizure frequency ranged from one seizure every 8 weeks to greater than 300 seizures per month. Seizures tended to cluster in eight patients (50%) and were predominantly nocturnal in six (37.5%). Only one patient reported catamenial exacerbation of seizures. Six patients (37.5%) had occasional or frequent secondary generalized seizures until the time of surgery, and four (25%) had a remote history of tonic-clonic seizures. Nine patients described auras; five patients described fear or panic, two an "indescribable sensation," and one each noted speech arrest, "light-headedness," and focal sensory symptoms. All had complex partial seizures. Only two patients had a history of status epilepticus.

Seizures. Ictal behavior, summarized in table 2, was characterized by combining patients' subjective reports with nursing and video monitoring records. Average seizure duration was 68 seconds (range, 10 to 335 seconds), derived by averaging each patient's mean seizure duration (total of 60 seizures witnessed in hospital). Early motor manifestations (within the first few seconds) were the most common ictal behavior (15/16 patients), and vocalization (phonation of any kind) was also prominent (10/16 patients). Vocalization occurred in patients with both dominant and nondominant foci. Seven patients had bilateral complex automatisms (irregular, repetitive or nonrepetitive, purposeful or nonpurposeful movements of the limbs). No oral or alimentary automatisms were noted. No patients reported prolonged postictal periods after their complex partial seizures; five patients reported short periods (less than 5 minutes), and 11 patients had no significant postictal confusion but returned to baseline mental function rapidly.

Scalp EEG. The interictal scalp EEG results are presented individually in table 1. Interictal spikes were present in 13 of 16 patients (81%). They were restricted to the epileptogenic frontal lobe in only four patients. Of the remaining 12 patients, (1) one patient had bilateral asynchronous frontal spikes; (2) four patients had generalized spike and wave discharges (two accompanied by ipsilateral focal frontal spikes, one accompanied by bilateral asynchronous frontal spikes, and one unaccompanied by focal frontal spikes); (3) one patient had multilobar interictal spike foci in the hemisphere ipsilateral to the affected frontal lobe; (4) three patients had temporal lobe spikes (two in the temporal lobe ipsilateral to the frontal foci, one with bilateral temporal spikes); and (5) three patients had no spikes in the interictal scalp EEG. The generalized spike and wave discharges were associated with unresponsiveness or "pseudoabsence" attacks in only one patient, who had a frontopolar lesion; all other patients with the generalized spike and wave discharges had complex partial seizures.

Ictal scalp/sphenoidal EEG was recorded in 15 patients (table 1). (1) Five patients had localized frontal seizure onsets, but seizure onset was correctly localized in only three patients. In two patients, seizures were misleadingly localized; in patient 2, seizures were proven to originate in the opposite frontal lobe with intracranial EEG, and in patient 3, intracranial EEG showed frontal seizures arising independently from each frontal lobe. (2) Two patients had ictal onset ipsilateral to the affected frontal lobe, but the discharges were widespread in the hemisphere. (3) In three patients, the scalp EEG showed bilateral changes when each seizure began. (4) In five patients, all or some complex partial seizures had no EEG ictal correlate. Three patients had no obvious ictal discharge during all their complex partial seizures; the EEG showed only muscle artifact and general attenuation of amplitude. In two patients, most complex partial seizures also were without obvious electrographic correlate, but some seizures showed bilateral ictal changes at seizure onset. Both patients had been pseudoseizure suspects.

Intracranial EEG. Chronic intracranial EEG was performed before surgery in seven patients (table 1). In five patients, ictal intracranial EEGs helped identify the area to be resected; in two of these cases, EEG identified an area later demonstrated to contain a tumor (hamartoma, ganglioglioma) that had not been seen with MRI. In one patient, intracranial EEG demonstrated independent frontal seizures; this patient has responded favorably to anterior corpus callosotomy alone. In one patient (patient 4), intracranial EEG was not definitive, but resection was offered because of strongly lateralized extracranial EEG findings.

Electrocorticography was performed in 13 of the 14 patients who had frontal resections to help define the resection margins. Corticographic recordings were evaluated for spikes and nonepileptiform features, such as focal slowing, focal loss of background fast activity, or localized failure of medication-induced beta induction, when determining the extent of resection. Corticography was performed repeatedly during these operations, often leading to further resection. For patients with persistent spikes in whom further resection was thought inadvisable, anterior corpus callosum section (3), transverse subpial gyral sections (1), or cortical isolation (deafferentation) (1) was carried out. The patient without electrocorticography at the time of resection had extensive preresection subdural recordings and a frontal resection of all cortex anterior to the motor strip along with an anterior callosotomy. Only nine patients had postexcisional electrocorticography, too few to draw definitive conclusions about its usefulness in this population. Two of four patients with postexcision spikes became seizure-free, whereas four of five patients without spikes became seizure-free.

Chronic intracranial EEG was performed after focal resective surgery in six patients and after anterior corpus callosotomy in four patients. These recordings provided information that led to further surgery in two patients in each group. In one case, intracranial EEG lateralized the focus only after anterior corpus callosotomy had been done (patient 6).

MRI. Brain MRI revealed a frontal lobe lesion in nine of 16 patients. An additional two patients with normal high-quality MRIs (1.5-tesla machine) proved to have histopathologic lesions that were found at surgery (hamartoma, ganglioglioma).

Pathology. Histopathology was available in 14 patients (table 1). Six patients had discrete tumors (hamartoma, ganglioglioma, epidermoid, low-grade glioma, venous angioma), and three patients with a remote history of excised tumors had gliosis in the excised specimen. Three additional patients had gliosis, and only one of these had a known antecedent event (encephalitis).

Outcome. Nearly all patients improved after surgery. Fifteen patients had at least 1 year of follow-up (mean, 46 months; range, 12 to 79 months). Of these 15, 10 patients (67%) had a class 1 outcome, one patient (7%) had a class 2 outcome (exclusively nocturnal seizures), three patients (20%) had a class 3 outcome, and one patient (7%) had a class 4 outcome.

We performed several analyses on the 15 patients with at least 1 year of follow-up to examine preoperative predictors of outcome. Patients who became seizure-free after surgery had a lower preoperative seizure frequency (16.0 + 19.6 seizures per month) than patients who had persistent seizures after surgery (102.0 + 114 seizures per month) (Mann-Whitney test, Z = 2.45, p < 0.05). Seizures tended to begin at a later age in patients who were seizure-free after surgery (17.3 + 11.8 years) than in those who were not seizure-free after surgery (7.0 + 3.2 years) (p < 0.10). Preoperative MRI did not correlate with outcome. Patients with preoperative frontal lobe MRI abnormalities were as likely to become seizure-free (5/8) as those without frontal lobe MRI abnormalities (4/7).

There was no major persistent neurologic morbidity (hemianopsia, hemiplegia, hemisensory deficit, aphasia) or mortality. One patient had a substantial flattening of affect and personality change. Three patients had transient (1 to 2 months) hemiparesis presumably related to excision in the supplementary motor area (SMA).

Discussion. Frontal lobe seizures remain difficult to diagnose due to their protean behavioral manifestations and the relative insensitivity of conventional diagnostic tools. Consequently, it is desirable to further define their characteristics, particularly those that are shared by patients with that diagnosis. The results presented here, although indicating some common features, lead one to conclude that there is a striking heterogeneity in clinical expression of frontal lobe disease. Nonetheless, some attributes might help distinguish FLE from temporal lobe epilepsy in many patients.

Most frontal lobe seizures caused early bilateral movements and most had minimal postictal confusional states, confirming earlier reports ^{[3,5,6,8,13,14]}. Notably absent were the oroalimentary automatisms that are so common in epilepsy of temporal lobe origin ^[9]. Although vocalization was common, it had no lateralizing or localizing value. Complex partial seizures lasted an average of 68 seconds in our patients, more than twice that reported previously, although they were still much shorter than the typical temporal lobe complex partial seizure, which may last several minutes ^[3]. As in temporal lobe epilepsy, most of our patients reported an aura, in contrast with other reports ^[3]. Five patients described an aura of fear or panic, often reported with temporal lobe seizures; cingulate cortex (mesial frontal), a site that can produce fear, ^[13] was implicated in three of these individuals. Although Williamson et al ^[3] described frontal lobe seizures as being primarily nocturnal and occurring in clusters, less than half our patients displayed these characteristics, and a minority had secondarily generalized seizures. We also found a wide variation in the seizure frequency of patients suffering from FLE. Of note, six patients (38%) had more than 30 seizures per month, a rare phenomenon in temporal lobe epilepsy ^[15]. Although age of seizure onset also varied, a relatively small proportion of these patients (25%) had an early risk factor for epilepsy or seizures beginning by age 5. This contrasts with temporal lobe epilepsy patients, in whom more than three-fourths have an early risk for epilepsy or seizures beginning by age 5 ^[15]. Hence, features suggestive of FLE consisted of a relatively high monthly seizure frequency, older age of onset, and seizures lasting less than approximately 1 minute.

A number of investigators have characterized the semiology of seizures from specific frontal lobe loci. Among the best described are those emanating from supplementary motor cortex. SMA seizures are usually brief, cluster at night, and may be associated with vocalization or preservation of consciousness ^[9,14]. Speech arrest, contraversive head movement and eye deviation, contralateral arm abduction, and external rotation and flexion at the elbow are believed to be virtually pathognomonic for an SMA focus. Seizures arising from adjacent mesial frontal structures, such as cingulate cortex, may produce prominent emotional and autonomic symptoms with an aura of panic or fear and evidence of increased sympathetic tone ^[13]. These symptoms are not specific for cingulate seizures, as temporal lobe seizures are also frequently associated with panic and autonomic discharge.

The semiology of orbitofrontal seizures is less well characterized. These may remain silent until they spread to adjacent cingulate or insular regions, which may cause associated autonomic signs, olfactory hallucinations, or oroalimentary automatisms ^[16]. Dorsolateral seizures are also not well characterized in the literature, although they have been associated with tonic posturing of the head and upper extremities, contraversive head and eye movements, and occasionally clonic facial contractions ^[17]. It is unclear whether seizures of frontopolar origin have distinct clinical manifestations; these may be caused by ictal spread to adjacent areas. They have been associated with pseudoabsence, loss of postural tone, and secondary generalization ^[9,14]. One of our patients fits this description.

Unfortunately, the relation between clinical manifestations and the precise site of origin of a particular seizure remains problematic. Many patients have rather sizable areas of cortex implicated in seizure initiation, and once begun, frontal lobe seizures are rarely confined to a small volume of cortex. Consequently, many ictal manifestations may result from seizure spread either to adjacent or distant cortex. In many of our patients, ictal behavior, although suggestive of frontal lobe disease, was difficult to correlate with any specific frontal lobe location.

The scalp EEG was abnormal in most of our patients. When lateralized interictal spikes were present, they usually indicated which hemisphere was abnormal, but these often were not restricted to the frontal lobe. Indeed, several patients had spikes confined to the temporal lobes. Because spikes also appeared in the hemisphere contralateral to the affected frontal lobe, occasionally with prominence, excessive reliance cannot be placed on these interictal discharges. In accordance with earlier reports, ^[10] we occasionally noted generalized spike wave or secondary bilateral synchrony; this occurred considerably more often than in a population with refractory temporal lobe epilepsy, in whom bisynchronous spikes are rare ^[12,18]. The ictal EEG recorded with scalp electrodes less frequently offered useful localizing information, occasionally defining a unilateral frontal focus, but more often showing bihemispheric changes without localizing information. As noted previously, ^[3] the scalp EEG may not contain clear-cut ictal abnormalities during seizures; this occurred in five patients in this series. This contrasts with temporal lobe epilepsy, in which the scalp EEG of complex partial seizures nearly always displays ictal changes ^[12,18]. Despite the limitations of scalp EEG, it provided definitive localizing information in one patient (patient 4) in whom the intracranial EEG could not define a resectable lesion.

Chronic intracranial EEG generally aided in planning surgery, but its sensitivity and specificity are limited. For some patients, it clearly provided essential data, but resecting an area from which well-localized seizures originated did not always guarantee seizure relief. Moreover, inconclusive intracranial EEG was still followed by a successful unilateral resection in one patient. Acute intraoperative electrocorticography also aided in defining tissue for resection, and we advise consideration of both epileptiform and nonepileptiform features to optimize the value of this technique. At present, we recommend bilateral frontal and temporal lobe sampling in FLE patients, with extensive dorsolateral, mesial, and basal frontal coverage, while placing more electrodes in the suspect region (determined by prior noninvasive studies). The lack of reliability of scalp EEG mandates the broad sampling. How to best integrate the intracranial EEG with other data remains to be determined, and this task will require a larger number of patients.

MRI is an especially valuable tool. Although often normal in FLE, ^[19] it usually defines lesions well when they are present. However, MRI is not infallible, and two of our patients with normal MRIs later proved to have lesions that were located by chronic intracranial EEG monitoring and delineated at surgery. Moreover, having a discrete preoperative MRI abnormality did not favorably influence prognosis.

Two surgical procedures are commonly used in the management of FLE: cortical resection and corpus callosotomy. Resection of the epileptogenic lesion is desirable, for it offers the best chance of a cure. However, sometimes the extent or location of the lesion cannot be defined, and at other times, it cannot be safely excised. In this circumstance, anterior corpus callosotomy may ameliorate seizures ^[20-26]. The effectiveness of this procedure probably relates to the role that the corpus callosum plays in seizure spread and, perhaps, in modulating cortical excitability. Because anterior corpus callosotomy alone improves outcome, we used it in conjunction with a frontal resection, hoping to optimize results when a combined procedure was technically feasible. We thought this potentially worthwhile because patients with FLE often respond incompletely to focal resection ^{[10,20,24,25]}. Our small series size precludes definitive conclusions regarding the relationship between postoperative seizure control and type of surgical procedure, and this bears further investigation. Some preoperative clinical features, including preoperative seizure frequency and age of seizure onset, may also correlate with outcome, but these defy ready explanation and require added study. Although the absence of a tumor may portend a poor prognosis ^[10] (chiefly because of resultant uncertainties in localization), improvements in other diagnostic modalities may sufficiently improve precision and therefore compensate for the lack of an obvious structural lesion.

The frequency of the FLEs still remains open to question. In our center, they constituted approximately 6% of the total number of patients offered surgery. This percentage applies only to those with intractable epilepsy, and therefore may not reflect its prevalence in the general population of people with epilepsy. The low frequency of this condition in our series could arise from several causes, including referral bias, difficulty of diagnosis, strictness of inclusion criteria (we required refractory epilepsy to enter the database, not just a history of seizures in the setting of a tumor), or lower prevalence than previously estimated ^[1]. Epidemiologic studies are needed to address this issue.

The diagnosis and surgical management of FLE remain beset by difficulties, but compared with historical reports, progress is being made. Although a heterogeneous condition, aggressive investigation and intervention is warranted because treatment often is successful. More work must be done to define the role of other neuroimaging techniques, ^[27,28] and efforts should be increased to improve the prediction of outcome. Reports about FLE ideally should include patients with proven frontal lobe disease, proven by response to surgery, because reliance on one aspect of the evaluation (eg, electrophysiology, clinical behavior) for diagnosis can be misleading.

Acknowledgment

The authors gratefully acknowledge the assistance of C. Plummer in retrieval of data for this review.