After obtaining approval from the Seoul National University Hospital (SNUH) review board, from a prospective database we retrospectively identified 153 consecutive patients without MRI-detectable lesions who underwent epileptic surgery at the SNUH between September 1995 and March 2011. As we applied a new patient selection strategy after July 2002, we classified these patients into two groups according to their surgery date. Before July 2002, we considered surgical treatment of MRI-negative neocortical epilepsy even when they had at least one localizing result of presurgical studies (ictal scalp EEG, interictal scalp EEG, FDG-PET, and ictal-interictal SPECT) with the consideration of seizure semiology, while our new patient selection strategy required concordant results of two or more presurgical studies. Although we tried to apply the new strategy, a few individuals who had lower concordance were included in this new group based on their invasive study results, because the final decision to take surgical treatment was usually made by the patient or their care-giver after the explanation of the possibility of successful surgery based on the results of presurgical studies. So we excluded 11 patients with low concordance (0–1) from the new group for comparison between old and new groups. The old group consisted of 89 patients and the new one consisted of 53 patients. Concordance of presurgical studies was defined as the results of ictal/interictal EEG and functional neuroimaging were localized to the same epileptogenic lobe. We included only patients who received focal resection and had a follow-up period of at least 2 years. The mean follow-up period was 11.4 ± 5.16 years for the old group and 4.2 ± 1.65 years for the new group. We excluded patients with epileptic surgery other than focal resection, such as functional hemispherectomy or corpus callosotomy. Patients with a medial temporal ictal onset zone were excluded.

A multidisciplinary presurgical evaluation was performed. All patients underwent a thorough history taking and physical/neurological examinations. All patients underwent brain MRI using an epilepsy protocol as described previously,10 MRI was performed on 1.5-T or, since 2003, 3-T systems. Our imaging protocol included 3 mm T2-weighted/fluid-attenuated inversion recovery oblique coronal images and 1.5 mm T1-weighted three-dimensional magnetization-prepared rapid acquisition with gradient-echo sequences. An experienced neuroradiologist who was blinded to other presurgical evaluation findings viewed the MRI first. The secondary interpretations were performed at a multidisciplinary case conference based on data from video-EEG monitoring, FDG-PET, and ictal-interictal SPECT, if available.

Continuous video-EEG monitoring was performed in all patients, to capture at least three habitual seizures. The international 10–20 system and additional anterior temporal electrodes were used for scalp EEG monitoring. Interictal epileptiform discharges and ictal-onset rhythms were classified as localizing if they were confined to the electrodes of the epileptogenic lobe. According to the ictal or postictal phenomenon including subjective symptoms during video-EEG monitoring and the stereotypic history from medical records, we classified their semiology to localized, lateralized or not helpful.

If possible, functional neuroimaging studies were performed as described previously.10 FDG-PET during the interictal period (no seizures for more than 24 h) was performed in 143 patients. FDG-PET images were assessed visually and using statistical parametric mapping (SPM) analyses. Ictal and interictal SPECT was performed in 101 patients. Side-by-side visual analysis of interictal and ictal images and a subtraction method were performed. The results of functional neuroimaging studies were defined as localizing if the predominant hypometabolic zone (FDG-PET) or the predominant hyperperfusion area (ictal-interictal SPECT) was confined to the epileptogenic lobe.

After these noninvasive studies, extraoperative intracranial monitoring was performed in all patients to define the resection margin. We used combinations of subdural grids and strips. The placement of the grids and strips was determined by the results of noninvasive studies (video-scalp EEG monitoring, FDG-PET, and ictal-interictal SPECT). We categorized the distribution of seizure onset as focal (involving <5 adjacent electrodes), regional (involving <5 adjacent electrodes), or widespread (involving >20 adjacent electrodes). The ictal onset zone of extraoperative intracranial monitoring was defined as the area with the first sustained rhythmic change in EEG that was differentiated from the background and interictal waves. When necessary, preoperative and intraoperative functional mapping and intraoperative electrocorticography were performed.

The ictal onset zone and the area with frequent interictal spikes accompanied by persistent pathological delta slowing of extra-operative intracranial monitoring were included in the resection area (epileptogenic lobe), if possible. The resection margin was tailored to the location of the eloquent cortex. All operations were performed by one neurosurgeon (C. K. Chung). Tissue sections from operation were immersed and fixed in 10% buffered formalin, embedded in paraffin, and stained with hematoxylin and eosin, cresyl violet, and Bielschowsky stain.

Outcome status was assessed via outpatient clinical interviews. The outcome was stratified according to Engel’s classification: class 1, seizure free; class 2, >90% reduction; class 3, >50% reduction; and class 4, <50% reduction or unchanged. The surgical outcomes and follow-up duration were evaluated according to the most recent medical record. The recent outcome of 89 patients in old group did not change significantly as compared with previously reported outcome (mean Engel’s classification 2.17 vs. 2.22, p = 0.759). Six patients in old group were improved and five patients were aggravated according to the most recent medical record as compared with previously reported outcome.

Data storage and statistical analyses were performed using SPSS software (19.0; IBM SPSS Statistics for Windows, Armonk, NY, USA). Student’s t test was used for the analysis of continuous variables. The χ² test or Fisher’s exact test of independence was used for univariate analyses. A stepwise multiple linear regression analysis was performed for multivariate analyses. Statistical significance was set at p < 0.05.

There were no differences in sex, age at surgery, onset of non-febrile seizure, and disease duration between the old and the new groups (Table 1). Among the 153 patients included in the study, the mean age at surgery was 27.2 years (range, 7–56 years) and 55.6% were male. The number of patients with localized or lateralized seizure semiology was not different between the two groups. After application of the new strategy, resection of neocortical temporal lobe was increased (34.8% vs 62.3%; p = 0.002). There were 53 cases of cortical dysplasia in the old group and 41 cases in the new group. The percentage of cortical dysplasia was not different between two groups.

The number of presurgical evaluations was not different between patients of the new group and the old group (Table 1). However the number of patients with FDG-PET was significantly high in the new group (88.8% vs. 100%; p = 0.014). The mean concordance of pre-surgical evaluations (range, 0–4) and the number of patients with concordance between two or more evaluations were higher in the new group compared with the old group. The concordance among presurgical evaluations for extratemporal resection increased after implementation of new strategy.

Surgical outcome was improved after the application of the new strategy (Table 2). In the old group, 73 (82%) patients had a good surgical outcome (EI-III) and 42 (47.2%) patients were seizure free. Compared with the old group, a higher percentage of patients in the new group had a good surgical outcome (52 patients, 98.1%; p = 0.003), including 40 patients who were seizure free (75.5%; p = 0.001).

The univariate analysis revealed that sex, age at surgery, age at onset of nonfebrile seizure, duration of epilepsy, and localized/lateralized semiology were not significant prognostic factors of outcome (Table 3). In terms of presurgical evaluation, localizing interictal scalp EEG (p = 0.019), FDG-PET (p = 0.000), ictal-interictal SPECT (p = 0.016), and ictal scalp EEG (p = 0.001) were associated with a seizure-free outcome. Concordance between two or more modalities was also related to seizure-free outcome (p = 0.000). The parameters of invasive studies, including onset pattern and onset frequency of ictal discharge, were not related to surgical outcome. Neocortical temporal location of surgical resection was a good prognostic factor (p = 0.002). Binominal multiple logistic regression identified only two or more concordant modalities (p = 0.007) and localizing FDG-PET (p = 0.018) as independent prognostic factors of seizure-free outcome (Table 4).

Localizing value of individual presurgical modalities according to the resected lobe was estimated (Table 5). The localizing value of FDG-PET was greatest in neocortical TLE. The predictive value of noninvasive presurgical modalities was estimated (Supplementary Table 1). FDG-PET showed higher positive predictive value (0.72) rather than other modalities.