Prognostic Factors in Neocortical Epilepsy Surgery: Multivariate Analysis
Address correspondence and reprint requests to Dr. S.K. Lee at Department of Neurology, Seoul National University Hospital, 28 Yonkeun dong, Chongno ku, Seoul 110-744, Korea. E-mail: firstname.lastname@example.org
Summary: Purpose: Defining prognostic factors for neocortical epilepsy surgery is important for the identification of ideal candidates and for predicting the prognosis of individual patients. We use multivariate analysis to identify favorable prognostic factors for neocortical epilepsy surgery.
Methods: One hundred ninety-three neocortical epilepsy patients, including 91 without focal lesions on MRI, were included. Sixty-one had frontal lobe epilepsy (FLE), 80 had neocortical temporal lobe epilepsy (nTLE), 21 had parietal lobe epilepsy (PLE), and 22 had occipital lobe epilepsy (OLE). The primary outcome variable was patient status ≥2 years after surgery (i.e., seizure free or not). Clinical characteristics and the recent presurgical diagnostic modalities were considered as probable prognostic factors. Univariate and standard multiple logistic regression analyses were used to identify favorable prognostic factors.
Results: The seizure-free rate was 57.5%. By univariate analysis, a focal lesion on MRI, localized ictal onset on surface EEG, epilepsies other than FLE, localized hypometabolism on fluorodeoxyglucose–positron emission tomography (FDG-PET), and pathologies other than cortical dysplasia were significantly associated with a seizure-free outcome (p < 0.05). Multivariate analysis revealed that a focal lesion on MRI (p = 0.003), correct localization by FDG-PET (p = 0.007), and localized ictal onset on EEG (p = 0.01) were independent predictors of a good outcome.
Conclusions: The presence of a focal lesion on MRI, correct localized hypometabolism on FDG-PET, or localized ictal rhythms on EEG were identified as predictors of a seizure-free outcome. Our results suggest that these findings allow the selection of better candidates for neocortical epilepsy surgery.
Epilepsy surgery is now widely accepted for the management of medically intractable localization-related epilepsy (1,2). The ideal goal of epilepsy surgery is to achieve complete seizure control in medically intractable patients (3). Extratemporal resections were common procedures in the past (4). However, in recent epilepsy surgery series, these procedures were used in only a small proportion of cases, whereas ≤80% of procedures in current surgical series involve the temporal lobe (4). Experience at large surgical centers indicates that patients with neocortical epilepsies constitute 20–40% of subjects considered for surgery (5–7). Patients with medial temporal lobe epilepsy (TLE) or epilepsy with a focal circumscribed epileptogenic lesion on MRI may be highly preferred candidates for epilepsy surgery, as these diagnoses are good predictors of favorable surgical outcomes (8–10).
The key to resective surgery success for such patients is the accurate localization of the region of seizure origin. However, the evaluation of patients with neocortical epilepsy requires multimodal diagnostic approaches and invasive studies. The number of potential predictors of seizure remission has grown as additional measures of brain structure and function have become available. Thus methods of identifying prognostic factors that can be used to choose ideal candidates for surgery and that also can be used to predict individual prognoses in cases of neocortical epilepsy are very important.
However, the results of work conducted to date to identify valuable prognosis predictors for neocortical epilepsy surgery are limited, and such studies have usually been performed on only small numbers of patients. Furthermore, multivariate analysis has been used in data analyses, because of the limited numbers of patients studied. The objective of the present study was to identify prognostic factors by using univariate and multivariate analyses, including data of multimodal neuroimaging studies, which would serve as reliable predictors of patient benefit after surgery for neocortical epilepsy.
We evaluated retrospectively a group of 193 consecutive neocortical epilepsy patients who had undergone focal neocortical resection between October 1994 and October 2001 at Seoul National University Comprehensive Epilepsy Center. A diagnosis of neocortical epilepsy was made by using the following criteria: (a) a focal neocortical lesion was present on MRI and a compatible ictal onset was identified during video-EEG monitoring; or (b) the brain MRI was normal, but an invasive study with intracranial electrodes confirmed neocortical ictal onset. All patients had intractable epilepsy despite proper anticonvulsant medication (AEDs). The duration of mean follow-up after operation was 43.7 ± 15.4 months. In this study, we included only patients who received focal resection. Patients with functional hemispherectomy and corpus callosotomy were excluded, as were 13 patients lost to follow-up and three patients inoperable because of an eloquent cortex identified by invasive study.
Clinical parameters and presurgical evaluation
The clinical characteristics available for each subject included age at surgery, the duration of epilepsy, age at afebrile seizure onset, monthly seizure frequency, presence of a preceding possible cause for epilepsy, existence of aura or secondarily generalized tonic–clonic seizures (2GTCS), and the presence of a preoperative neurologic deficit.
A multidisciplinary presurgical evaluation was performed, which included a complete neurologic examination, brain MRI with TLE protocol (as previously described) (11), long-term video-EEG monitoring, fluorodeoxyglucose–positron emission tomography (FDG-PET), intracarotid amobarbital test, and ictal and interictal single-photon emission computed tomography (SPECT) if possible.
Evaluation of noninvasive studies
Interictal/ictal scalp EEGs were reviewed and classified by two epileptologists after they reached consensus. A localizing pattern of ictal-onset rhythm or interictal spikes was defined as a localized ictal rhythm or interictal spike pattern confined to the electrodes of an epileptogenic lobe or to two adjacent electrodes covering two adjacent lobes (for example, T6 and O2). An interictal pattern was classified as “localizing” if a cortical area corresponding to one of two adjacent electrodes was included in the surgically resected area. Even when localizing ictal rhythm started from two adjacent electrodes covering two adjacent lobes, invasive studies confirmed that ictal rhythm commenced predominantly from one lobe. We classified patients in this category as having unilobar epilepsy and performed subsequent resection of the lobe. If a patient had localized ictal rhythm in at least one ictal event among many seizures during monitoring, we classified this patient as having a localized ictal scalp EEG. However, if a patient showed multifocal localization patterns of ictal onset or only diffuse patterns, the patient was regarded as having a nonlocalized ictal EEG.
Methods for PET and SPECT image acquisition, as well as methods for statistical parametric mapping (SPM) analysis, have been described elsewhere (11,12). FDG-PET images were analyzed visually and by SPM, and ictal–interictal subtraction SPECT images were reviewed by an experienced physician who was unaware of patients' clinical histories or the results of other presurgical evaluations. Interictal and ictal SPECT images were also evaluated by side-by-side visual analysis. FDG-PET and SPECT results were defined as “focal” when the hyperperfusion area or the hypometabolic zone was concordant with the resected area.
Decision to perform surgery
The criteria used for surgical resection were as follows: the presence of a discrete lesion on MRI with compatible video-EEG monitoring, or of an ictal-onset zone confirmed by intracranial EEG. We also performed invasive monitoring on patients with a lesion shown by MRI if (a) the lesion was near eloquent cortex, or (b) the lesion might be cortical dysplasia, or (c) a definite discrepancy existed between the results of presurgical evaluations. In some other patients demonstrating lesions on MRI, we performed simple lesionectomy and marginectomy when the results of ictal scalp EEG and semiology were well correlated with the locations of the lesions.
Surgical resection procedures in patients with normal MRI scans followed three rules. First, the region showing clear ictal EEG onset before clinical ictal onset was included in the resection area. Second, the area with frequent interictal spikes accompanied by persistent pathologic delta slowings also was included, even though this area did not show initial ictal rhythm. Third, the area satisfying the first or second rule should not be an eloquent cortex.
Tissue sections from cortical resections were immersion fixed in 10% (vol/vol) buffered formalin, embedded in paraffin, and stained with Bielschowsky stain, cresyl violet, and hematoxylin and eosin. Diagnoses of pathologic cortical dysplasia were classified into mild, moderate, and severe, according to the system of Mischel et al. (13).
Only patients with a follow-up period of ≥2 years were included. Outcome status was determined by outpatient clinical interviews or by telephone interviews during the last year of follow-up. Patients were categorized as seizure free or not.
Student's t test was used for the analysis of continuous variables. These included age at operation, age at onset, epilepsy duration before surgery (defined by difference between age at operation and age at onset), and seizure frequency per month before surgery. The χ2 test or Fisher's exact test of independence was used for the univariate analysis of all categoric variables, which included sex, history of a preceding insult, presence of a aura, recorded preoperative deficit, presence of 2GTCS, location of epileptogenic zone, laterality of the epileptogenic hemisphere, focal interictal epileptiform discharge, demonstration of localized ictal scalp EEG, presence of a focal lesion on MRI, demonstration of focal hypometabolism on PET, and data indicating focal hyperperfusion on ictal SPECT compared with interictal SPECT. The resected area was viewed as the gold standard for the various presurgical imaging studies. The need for invasive study, the presence of early seizure after operation, and the pathology also were included in the outcome analysis. Univariate analysis was performed on these variables with respect to outcome ≥2 years after surgery. Data storage and statistical analysis were performed by using SPSS (SPSS, Inc., Chicago, IL, U.S.A.). Standard forward stepwise multiple logistic regression (significance level, <0.05), using the postsurgical outcome as a dependent variable, also was performed. We included independent variables with p values of ≤0.25 in the univariate analysis.
Of the patients, 61 had frontal lobe epilepsy (FLE), 80 had neocortical temporal lobe epilepsy (nTLE), 21 had parietal lobe epilepsy (PLE), 22 had occipital lobe epilepsy (OLE), and nine had multifocal epilepsy. The proportion of seizure-free patients was 57.5%≥2 years after surgery (Table 1). Postoperative follow-up duration did not differ between patients who were seizure free (42.4 ± 15.1 months) and those who had postoperative seizures (45.5 ± 15.6 months; t test, p = 0.18). Before surgery, all patients received interictal and ictal EEG and MRI, whereas PET was performed in 179, and ictal SPECT with interictal SPECT was carried out in 136 patients.
Table 1. Clinical characteristics and outcome after neocortical resection in 193 neocortical epilepsy patients; univariate analysis of prognostic factors for a seizure-free outcome for at least two years after surgery
|Age at operation (yr, mean ± SD)||26.4 ± 7.5||27.0 ± 8.4||0.64|
|Age at onset (yr, mean ± SD)||12.9 ± 7.1||13.6 ± 8.8||0.56|
|Duration of epilepsy (yr, mean ± SD)||13.5 ± 7.3||13.3 ± 6.8||0.88|
|Seizure frequency per month (mean ± SD)|| 19.0 ± 44.1|| 17.7 ± 44.3||0.84|
|Presence of 2GTCS||94||70||0.90|
|History of possible causesa||56||47||0.35|
|Preoperative neurologic deficit||13||12||0.55|
|Presence of aura||83||55||0.24|
|Side of surgery (right)||62||38||0.07|
|Interictal EEG, focal IED||43||25||0.24|
|Ictal EEG, localized ictal-onset rhythm||79||45||0.02|
|MRI, concordant focal lesion||69||33|| 0.003|
|PET, concordant focal hypometabolism (n = 179)||67/107||29/72|| 0.003|
|Ictal SPECT, concordant focal hyperperfusion (n = 136)||32/77||21/59||0.48|
|Early recurring seizure after operation (<1 wk)|| 9||12||0.15|
|Pathology, cortical dysplasia||75||62||0.02|
| FLE||24||37|| |
| nTLE||57||23|| |
| PLE||12|| 9|| 0.001|
| OLE||15|| 7|| |
| Multifocal|| 3|| 6|| |
By univariate analysis, age at surgery, age at onset of epilepsy, duration of epilepsy, gender, the presence of possible causes of epilepsy, preoperative neurologic deficit, the presence of 2GTCS, seizure frequency, lateralization of the epileptogenic hemisphere, presence of aura, and the need for invasive study were not significant prognostic factors of postsurgical outcome (Table 1). Conversely, the presence of a focal lesion on MRI, a localized ictal onset shown by surface EEG, and correct focal hypometabolism on FDG-PET were all identified as factors that were significant predictors of a postsurgical seizure-free outcome (Table 2). Pathology other than cortical dysplasia also was a good prognostic factor, and the location of epileptogenic foci was found to be related to surgical outcome. Postsurgical prognoses were found to be poor in patients with FLE and multifocal epilepsy. Early seizure recurrence during the immediate postoperative period (≤7 days after operation) did not affect a patient's long-term outcome (p = 0.15). Standard stepwise multiple logistic regression identified only a localized lesion on MRI (p = 0.004), focal hypometabolism on FDG-PET (p = 0.004), and localized ictal EEG (p = 0.009) as independent significant prognostic factors for a seizure-free outcome.
Table 2. Clinical characteristics and outcomes after neocortical resection in 193 neocortical epilepsy patients; multivariate analysis of prognostic factors for a seizure-free outcome for ≥2 years after surgery
|Lesion (+) on MRI||2.545||1.359–4.768||0.004|
|Localization by PET||2.494||1.349–4.610||0.004|
|Localization by ictal scalp EEG||2.368||1.237–4.530||0.009|
Previous studies have suggested that 55–70% of patients undergoing temporal resection and 30–50% of patients undergoing extratemporal resection achieve a completely seizure-free state (14). Extratemporal resections are known to be generally less successful than are temporal lobe resections (4,15,16). The results of the present study suggest that many neocortical epilepsy patients are likely to benefit from surgical treatment, and 57.5% of our patients were seizure free ≥2 years after surgery. Previous studies have usually categorized epilepsy patients as having either temporal or extratemporal epilepsy. However, this categorization is defective, as it fails to acknowledge that neocortical epilepsy may start from the lateral temporal lobe.
Positive prognostic factors for the surgical resolution of neocortical epilepsy are known to be the presence of a tumor, an abnormal MRI, EEG/MRI concordance, and extensive surgical resection. In the present study, multivariate analysis indicated that a focal lesion on MRI, focal hypometabolism on FDG-PET, and a localized ictal onset on EEG are significant positive prognostic factors. The resection of an epileptogenic lesion with an ictal-onset zone is recognized as the most important factor for a good surgical outcome (17–19). The majority of surgical series suggested that the identification of a specific lesion usually leads to a favorable postsurgical outcome (4,19–22). Here, we included 91 patients who did not show focal lesions by MRI, and our results also indicate that the identification of a structural lesion, by MRI, has positive prognostic value. An important reason for an unfavorable operative outcome in patients with nonlesional neocortical epilepsy is the inherent difficulty of identifying the epileptogenic zone (5).
The role of FDG-PET in presurgical evaluation has been reduced by recent advances in MRI. Some previous works reported the low yield of PET in neocortical epilepsies (23–25). However, FDG-PET remains an important clinical tool in patients, especially in patients with nonlesional neocortical epilepsy. A recent study involving 462 cases showed that 32% of normal MRIs were associated with abnormal PET imagings (26). FDG-PET also indicated hypometabolism in 12 of 13 FLE children with normal MRI who had pathologic diagnoses of microdysgenesis (27). Furthermore, the application of SPM to FDG-PET data may yield more information than that obtainable by visual analysis only (28). In our study, FDG-PET localized the epileptogenic lobe in 29 of 72 patients with nonlesional neocortical epilepsies. Moreover, FDG-PET localization was significantly related, by univariate and multivariate analysis, to a seizure-free surgical outcome. This difference between the present study and previous reports might be related to improved PET scanner resolution. However, we had used the same machine that was installed in our hospital in 1994 and applied SPM technique to all of our previous data. For these reasons, we could not prove such a trend in our series. Another possibility is that neuronal migration disorder, the most common pathology in our present study, may be detected by FDG-PET. Focal hypometabolism on FDG-PET has been reported in a significant proportion of children with cryptogenic infantile spasms (29). About 64% of these patients showed unifocal or multifocal hypometabolism on FDG-PET, and a majority of these metabolic abnormalities was believed to represent dysplastic lesions. In previous studies performed by our group, FDG-PET was found to localize the epileptogenic lobe in >40% of nonlesional neocortical epilepsy cases (30) and in four of five focal cortical dysplasia patients with a normal MRI data (31). The localizing value of FDG-PET is significantly higher in neocortical TLE than in other epilepsies, and thus the inclusion of many neocortical temporal epilepsy patients in our current series may have elevated the diagnostic value of FDG-PET in the present study.
In this study, both visual and SPM analysis of FDG-PEG imaging were performed in a complementary manner. SPM has been accepted as a standard analytic method in functional neuroimaging. We previously demonstrated that SPM analysis showed sensitivity similar to that afforded by visual assessment (12). SPM analysis was helpful in the localization of the epileptogenic lobe in some patients whose hypometabolic area was not clearly discernible visually. With data from some other patients, visual examination of FDG-PET images was more useful than was SPM analysis. We would thus describe visual assessment and the SPM technique as complementary. We also used both the subtraction technique with SPM and side-by-side visual analysis to interpret SPECT results. Although recent studies have demonstrated that the clinical utility of periictal SPECT can be enhanced by using subtraction images (32–35), the subtraction method is not always superior to the visual analysis. Actual interictal scans do not always represent the true interictal period, and this discrepancy may occur in as many as 5% of interictal scans, probably because of the injection during the subclinical ictal period (36). Furthermore, even the interictal injection of radiotracer several hours after an ictal episode cannot guarantee true interictal images (37). Late postictal hyperperfusion was found in more than half of the patients. This could explain the increased perfusion seen in the epileptogenic zone on interictal SPECT images. Therefore both the visual and subtraction analysis are necessary to interpret SPECT images properly, and the use of both techniques enhances the diagnostic accuracy, although the use of subtraction images may be superior to the visual analysis in many cases.
Contrary to previous reports indicating the value of subtraction SPECT in the localization of the epileptogenic lobe and in the prediction of surgical outcome (33,34,38), ictal SPECT was not a good predictor of surgical outcome in the present study. In previous work, two-lobar patterns such as parietooccipital or frontotemproal hyperperfusion were regarded as “localized,” as was the predominant one-lobar pattern. In this work, however, we classify two-lobar patterns as “lateralized” (thus we do not include this group in the “localized” class). When patients with lateralized lesions are included with those showing localized lesions, the diagnostic sensitivity of ictal SPECT is enhanced. The lateralizing pattern was the reflection of propagated ictal activity, which is often observed in extratemporal epilepsies (39–41). Besides the difference in the definition of localization in the present study, compared with previous studies, a fixed threshold of subtraction SPECT may also have affected the results (11). Adjustment of this threshold value may improve the diagnostic SPECT sensitivity.
In neocortical epilepsy, ictal scalp EEG has limitations as a diagnostic tool. Test data are frequently nonlocalizing, and sometimes false localizing if the focus is very circumscribed or located in the depths of a sulcus (42–47). A profound motion artifact at ictal onset is another obstacle to accurate EEG interpretation. However, despite a lack of mesial and inferior surface coverage, several studies have disclosed the value of scalp-recorded seizures in neocortical epilepsies. We previously reported that 62% of neocortical epilepsy patients had at least one localizable ictal EEG (48). Another report demonstrated that data from surface ictal EEG were adequately localizing in 72% of cases and found ictal recordings useful for the localization/lateralization of focal seizures more often in temporal than extratemporal epilepsy, with the exception of mesial frontal lobe epilepsy (49). Surgery also was found effective when a majority of scalp-recorded seizures arose from later-resected lobes (50). The present study, demonstrates the effectiveness of ictal scalp EEG tests for predicting surgical outcome and for localizing seizure foci. Moreover, localizing ictal EEG data may provide a clue for the effective placement of intracranial electrodes.
Cortical dysplasia was identified as a negative prognostic factor by univariate analysis, and compared with other CNS lesions, tumors are known to have a higher chance of seizure remission (51,52). In tumoral cases, epileptogenic lesions are more easily detected and represent a more circumscribed pathology, whereas cortical dysplasia has a tendency to be more widespread than revealed by MRI (53,54). The attainment of freedom from seizures in patients with neuromigrational defects is uncertain because of difficulties in defining the epileptogenic cortex (55).
In the present work, seizure-free rates after surgery, in patients with of FLE and multifocal epilepsies, were lower than those in subjects with other forms of epilepsy. The low seizure-free rate in patients with multifocal epilepsy may be explained by incomplete resection of all the epileptogenic zones. The relatively large surface area of the frontal lobe may hinder easy identification of the epileptogenic zone. Surgical limitations to complete resection, because of both eloquent areas and the difficulties associated with adequate intracranial sampling of the medial and orbitofrontal areas might also be related to a poor outcome in patients with FLE.