Relationship between preoperative hypometabolism and surgical outcome in neocortical epilepsy surgery

Authors

  • Chong H. Wong,

    1. Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    2. Department of Neurology, Westmead Hospital, Westmead, New South Wales, Australia
    Search for more papers by this author
  • Andrew Bleasel,

    1. Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    2. Department of Neurology, Westmead Hospital, Westmead, New South Wales, Australia
    Search for more papers by this author
  • Lingfeng Wen,

    1. Department of PET and Nuclear Medicine, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Stefan Eberl,

    1. Department of PET and Nuclear Medicine, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Karen Byth,

    1. Millennium Institute, Westmead, New South Wales, Australia
    Search for more papers by this author
  • Michael Fulham,

    1. Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    2. Department of PET and Nuclear Medicine, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    Search for more papers by this author
  • Ernest Somerville,

    1. Department of Neurology, Westmead Hospital, Westmead, New South Wales, Australia
    2. Department of Neurology, Prince of Wales Hospital, Randwick, New South Wales, Australia
    Search for more papers by this author
  • Armin Mohamed

    1. Faculty of Medicine, University of Sydney, Sydney, New South Wales, Australia
    2. Department of PET and Nuclear Medicine, Royal Prince Alfred Hospital, Camperdown, New South Wales, Australia
    Search for more papers by this author

Address correspondence to Armin Mohamed, Associate Professor, Department of PET and Nuclear Medicine, Royal Prince Alfred Hospital, Missenden Rd Camperdown, NSW 2050, Australia. E-mail: arminm@icn.usyd.edu.au

Summary

Purpose:  Fluorine-18-fluorodeoxyglucose–positron emission tomography (FDG-PET) hypometabolism has been used to localize the epileptogenic zone. However, glucose hypometabolism remote to the ictal focus is common and its relationship to surgical outcome has not been considered in many studies. We investigated the relationship between surgical outcome and FDG-PET hypometabolism topography in a large cohort of patients with neocortical epilepsy.

Methods:  We identified all patients (n = 68) who had interictal FDG-PET between 1994 and 2004 and who underwent resective epilepsy surgery with follow up for more than 2 years. The volumes of significant FDG-PET hypometabolism involving the resected epileptic focus and its surrounding regions (perifocal hypometabolism) and those distant to and not contiguous with the perifocal hypometabolism (remote hypometabolism) were determined statistically using Statistical Parametric Mapping (voxel threshold p = 0.01, extent threshold ≥250 voxels, uncorrected cluster-level significance p < 0.05) and were compared with magnetic resonance imaging (MRI) and clinical and demographic variables using a multiple logistic regression model to identify independent predictors of seizure outcome.

Key Findings:  Remote hypometabolism was present in 39 patients. Seizure freedom was 49% (19 of 39 patients) in patients with glucose hypometabolism remote from the epileptogenic zone compared to 90% (26 of 29 patients) in patients without remote hypometabolism. In 43 patients with an MRI-identified lesion, seizure freedom was 79% (34 of 43 patients). In patients with normal MRI, cortical dysplasia was the predominant pathologic substrate. Multiple logistic regression analysis identified a larger volume of significant remote hypometabolism (p < 0.005) and absence of a MRI-localized lesion (p = 0.006) as independent predictors of continued seizures after surgery.

Significance:  In patients with widespread glucose hypometabolism that is statistically significant when compared to controls, epilepsy surgery may not result in complete seizure freedom despite complete removal of the MRI-identified lesion. The volume of significant glucose hypometabolism remote to the ictal-onset zone may be an independent predictor of the success of epilepsy surgery.

Fluorine-18-fluorodeoxyglucose–positron emission tomography (FDG-PET) is often performed routinely during presurgical epilepsy evaluation. Several studies have indicated that the presence of localized FDG-PET hypometabolism is predictive of a seizure-free outcome after surgery (O’Brien et al., 2001; Lee et al., 2005; Yun et al., 2006). However, glucose hypometabolism remote to the ictal-onset zone is also common. Although the presence of remote glucose hypometabolism has been reported in mesial temporal lobe epilepsy (MTLE) to relate to a poor surgical outcome (Choi et al., 2003; Wong et al., 2010), its relevance in predicting surgical outcome in neocortical epilepsy is less clear (Swartz et al., 1998; Juhasz et al., 2001; Jeha et al., 2007). In this study, we quantified FDG-PET abnormalities in neocortical epilepsy and determined their significance to the prediction of postoperative seizure outcome.

Methods

The study was approved by the ethics committees of the Western and Central Sydney Area Health Services, Sydney, Australia. Sixty-eight patients with medically intractable neocortical epilepsy were identified from the epilepsy surgery databases at the Westmead and Royal Prince Alfred hospitals, who had preoperative interictal FDG-PET study between 1994 and 2004 at Royal Prince Alfred Hospital and had undergone epilepsy surgery and were monitored for at least 24 months after surgery. Patients with recurrent seizures because restricted resection was performed due to its proximity to eloquent cortex were not included (n = 5). Patients with MTLE, nonresective neurosurgery (corpus callosotomy and multiple subpial transections), hemispherectomy, or dual pathology (other visible lesions on magnetic resonance imaging (MRI), including hippocampal sclerosis) were excluded from the study. All patients had axial and coronal T2-weighted and T1-weighted images and coronal fluid-attenuated inversion-recovery (FLAIR) on either Siemens Magnetom Vision 1.5 Tesla scanner (Siemens AG, Erlangen, Germany) with three-dimensional T1-weighted magnetization-prepared rapid-acquisition gradient-echo (MPRAGE) sequences or General Electric Signa 1.5 Tesla scanner (GE Healthcare, Milwaukee, WI, U.S.A.) with three-dimensional T1-weighted spoiled gradient-recalled echo (SPGR) sequences.

Demographic data were collated from the patients’ records. Patients who had secondary generalized tonic–clonic seizures (SGTCS) in the 12 months preceding FDG-PET imaging or surgery were classified as patients with SGTCS. Epilepsy classification was performed by two epileptologists (AB and AM) blinded to the FDG-PET study results and was based on clinical data, MRI, and the seizures documented by video-electroencephalography (VEEG). The classification criteria were the following: (1) ictal EEG and semiology with concordant focal MRI abnormalities (n = 26) or (2) ictal-onset zone mapped by intracranial EEG recordings; ictal-onset zone was defined by intracranial electrodes showing an initial ictal rhythm, before clinical seizure onset, that is sustained and discernible from background activities (Kim et al., 2004; Lee et al., 2005). Intracranial EEG was used in the localization of the epileptogenic zone in cases with focal MRI lesions in proximity to eloquent cortex or with nonlocalizing EEG and semiology (n = 17), and in those with normal MRI (n = 25). Surgical resection was performed when the presence of a discrete lesion visible on MRI had compatible prolonged VEEG monitoring or intracranial EEG ictal onset. All patients with normal MRI had resection following confirmation of ictal onset by intracranial EEG.

Seizure outcome following surgery was evaluated by medical record review. In patients who had not attended clinic review in the preceding 12 months, a telephone interview was conducted. The last date of follow-up or censor date in this study was July 31, 2011. Good seizure outcome was defined as free of disabling seizures for at least 2 years after surgery (Engel class I seizure outcome). All patients (n = 23) who did not become seizure-free had reevaluation, and surgical resection was felt to be complete. Completeness of resection was determined by epileptologists, neuroradiologists, neurosurgeons, and neuropathologists in epilepsy surgery and epilepsy pathology meetings. Complete resection of the epileptogenic zone was defined as complete excision of the structural lesion (if detected by MRI) and regions exhibiting initial ictal rhythm and adjacent areas consistently involved in early propagation and regions with active interictal spiking or fast activity with consistent focality in regional proximity to the ictal-onset zone.

FDG-PET images were quantitatively analyzed by using Statistical Parametric Mapping (SPM) software (version SPM2; Wellcome Department of Cognitive Neurology, London, U.K.). For the purpose of analysis, the epileptogenic zone was assumed to be the resection site. Results of SPM analysis were considered correctly localized if >75% of the voxels from the most significant hypometabolic cluster (lowest p-value of all clusters) were concordant with the resected area. For visual identification of the epileptogenic lesion, two experienced PET physicians who were blind to clinical information interpreted FDG-PET images and selected the suspected focus based on visual criteria, such as size or intensity of the hypometabolic areas. If interpretation differed, a consensus opinion was reached after joint review. Visual and SPM analysis of FDG-PET was localizing in 50 patients and nonlocalizing in 12 patients. One patient was localized by visual analysis only and five patients were localized only with SPM analysis. Across all the subjects, agreement between visual and SPM analysis was excellent for localization of the epileptogenic region (kappa = 0.75, p < 0.001) and remote hypometabolism (kappa = 0.67, p < 0.001).

FDG-PET acquisition and SPM analysis methods have been described previously (Hooper et al., 1996; Mohamed et al., 2005; Wong et al., 2010). All FDG-PET studies were performed as outpatient. To ensure that the FDG-PET scan was not carried out periictally, VEEG monitoring was carried out for 30 min before and throughout the FDG-PET uptake and scanning period. For all subjects, approximately 400 MBq of FDG-PET was administered, and the FDG-PET scan was performed on an ECAT 951R PET scanner (CTI/Siemens, Knoxville, TN, U.S.A.). Both patients and controls had the regional metabolic rate of glucose hypometabolism estimated using a population-based input function calibrated with two arterialized venous blood sampling (Eberl et al., 1997). Each patient’s image was compared with a group of 16 healthy controls by using a two-sample t-test to detect any regional decrease in metabolism. Height threshold (individual voxel-level significance) was set at p ≤ 0.01 (z-score, 2.34). Only clusters with more than 250 contiguous voxels (voxel dimensions = 2 × 2 × 2 mm) or 2 cm3 in volume and a cluster significance level of p ≤ 0.05 (uncorrected) were considered significant. This threshold was defined using leave-one-out cross-validation procedure on FDG-PET images of normal healthy controls as was done previously (Signorini et al., 1999) (Data S1). Regions considered significant on SPM were overlaid onto spatially registered and normalized anatomic-labelled Montreal Neurological Institute (MNI) brain template with 96 regions of interest (Tzourio-Mazoyer et al., 2002). These regions were grouped into six paired anatomic areas for analysis. The anatomic areas on each side included frontal, parietal, temporal, occipital, basal ganglia, and thalamus. The hypometabolic cluster involving the epileptogenic area and contiguous adjacent regions was defined as perifocal hypometabolism (PH), whereas the hypometabolic clusters distant to and not contiguous with the PH area were defined as remote hypometabolism (RH, Fig. 1). We divided the glucose hypometabolism clusters into PH and RH based on the physical separation of the significant clusters, as defined objectively and statistically by SPM analysis. That division was used to investigate the clinical relevance of PH and RH. The extents of significant glucose hypometabolism in PH and RH were determined by counting the number of voxels.

Figure 1.


Two examples of remote hypometabolism. (A) A patient with lesional right occipital lobe epilepsy with severe hypometabolism in the epileptogenic right occipital lobe and remotely in bilateral temporal lobes. Histopathology showed ganglioglioma and patient was not seizure-free (Engel class II) after surgery. (B) A patient with lesional right frontal lobe epilepsy showing significant glucose hypometabolism in the epileptogenic right orbitofrontal gyrus and a region of remote hypometabolism in the right inferior parietal gyrus. Histopathology showed type 2b cortical dysplasia and patient has remained seizure-free for 5 years after surgery.

SPSS for Windows (version 15; SPSS Inc, Chicago, IL, U.S.A.) was used to analyze the data. Two-tailed tests with a significance level of 5% were used throughout. The Spearman rank correlation was used to quantify the association between the extent of PH and RH and the age of seizure onset and the duration of epilepsy prior to undergoing FDG-PET. The influences of SGTCS and histopathology findings on the extent of hypometabolism were examined using Mann-Whitney tests.

To examine the clinical significance of SPM-defined PH and RH, univariable logistic regression analysis was used to assess the influence of gender, age of seizure onset, duration of epilepsy, presence of SGTCS, presence of MRI lesion, age at surgery, type of pathology and extents of PH and RH on seizure outcome. Multiple logistic regression analysis with backward stepwise variable selection was performed to identify the independent predictors of seizure freedom. Potential predictors entered into the model were those that were significant at p < 0.1 in the univariable analysis, together with seizure type and its interaction with each of these variables. Odds ratios and their 95% confidence intervals were used to quantify the degree of association.

Results

Demographic data

The study included 68 patients (33 female; median age 25 years old, 25–75% interquartile range [IQR] 15–37). Twenty-two of the patients had frontal lobe epilepsy (FLE), 24 had neocortical temporal lobe epilepsy (NTLE), 7 had parietal lobe epilepsy (PLE), and 15 had occipital lobe epilepsy (OLE). In this cohort, four patients had a second resection because the initial resection was felt to be incomplete. Forty-five patients (66.2%) achieved good outcome (median follow-up period, 9.7 years; IQR 7.9–11.2 years).

Histologic examination revealed neoplasm (n = 19, with three, low-grade astrocytoma; four, dysembryoplastic neuroepithelial tumor; six, ganglioglioma; and six, oligodendroglioma), cortical dysplasia (16, type 1a; 10, type 1b; 6, type 2a; and 4, type 2b), gliosis (n = 6), and vascular abnormality (n = 7, Table 1).

Table 1.   Histopathology and the corresponding epilepsy type according to seizure outcome
PathologyEpilepsyType 1a CDType 1b CDType 2a CDType 2b CDVascular abnormalityGliosisAstrocytomaGangliogliomaDNETOligodendroglioma
  1. () indicates normal MRI.

  2. DNET, dysembryoplastic neuroepithelial tumor; NTLE, neocortical temporal lobe epilepsy; FLE, frontal lobe epilepsy; PLE, parietal lobe epilepsy; OLE, occipital lobe epilepsy;

Good outcome
 FLE(2)103(1)(2)0212
 NTLE3(2)20311202
 PLE1001200000
 OLE   1 (1)   2 (2)   2 (1)0010010
Poor outcome
 FLE(1)(2)(1)0001102
 NTLE(2)(1)001(1)(1)020
 PLE(2)000010000
 OLE(3)000000100
 Total161064763646

Predictors of postoperative seizure outcome

Univariable analysis of postoperative seizure outcome showed that the presence of a localized lesion on MRI was statistically significant for a good outcome (Table 2, p = 0.007). Patients with a lower RH volume were more likely to be seizure-free (p < 0.005) than were patients with a high RH volume. Seizure freedom was achieved in 26 (90%) of 29 patients without RH, in 15 (56%) of 27 patients with RH volume less than half that of PH, and in only 4 (33%) of 12 patients where RH volume was at least half that of PH. Presence of SGTCS was significantly associated with poor outcome (p < 0.005), whereas age of seizure onset, duration of epilepsy prior to surgery, and PH volume were not significantly related to seizure outcome (Table 2). There was no significant association between the histologic findings and seizure outcome (p = 0.56). Patients with neoplasm and cortical dysplasia were also analyzed separately and showed no significant association for seizure outcome. Although all patients with type 2b cortical dysplasia (n = 4) in the study were seizure-free after surgery, the numbers were too small to draw conclusions. The duration of follow-up in patients that have remained seizure-free and have relapsed were not statistically different (p = 0.54).

Table 2.   Clinical characteristics and seizure outcomes
 Good outcomePoor outcomep-Value
  1. Good outcome, Engel class 1; NTLE, neocortical temporal lobe epilepsy; FLE, frontal lobe epilepsy; PLE, parietal lobe epilepsy; OLE, occipital lobe epilepsy; (M/F), (male/female); (Y/N), (Yes/No); SGTCS, secondarily generalized tonic–clonic seizures; PH, perifocal glucose hypometabolism; RH, remote glucose hypometabolism; IQR, interquartile range, 25–75%.

  2. a Localizing FDG-PET by SPM analysis.

  3. b Number of significant voxels.

Epilepsy type
 FLE/NTLE/PLE/OLE14/16/4/118/8/3/40.88
Gender (M/F)24/2111/120.80
Age of onset (median; IQR, years)11; 6–1716; 5–210.17
SGTCS (Y/N)14/3117/60.001
Localized lesion on MRI (Y/N)34/119/140.007
Localizing FDG-PETa (Y/N)37/818/50.75
Volume of PHb (median; IQR, voxel)1,222; 425–2,3581,410; 543–3,1810.53
Volume of RHb (median; IQR, voxel)0; 0–420852; 446–1,195<0.005
Duration of epilepsy (prior to surgery) (median; IQR, years)11; 7–1714; 9–230.29
Duration of follow up (median; IQR, years)9.8: 8.0–13.2 9.4: 7.0–10.70.54

The presence of a localized lesion in MRI, absence of SGTCS, and the RH volume were entered as candidate variables into a multiple logistic regression model, together with epilepsy type and its interaction with each of these variables. None of the two-way interactions between epilepsy type and the selected potential risk factors was significant. After adjustments for the type of epilepsy and using stepwise backward selection, the independent predictors for the continuation of seizures following surgery were the absence of a localized lesion on MRI (p = 0.006, odds ratio [OR] 6.59, 95% confidence interval [95% CI] 1.71–25.43) and the volume of RH (p < 0.005, OR/100 voxel increase 1.30, 95% CI 1.12–1.50).

Clinical factors that affect the extent of PH and RH

No significant association was found between the types of epilepsy, age of seizure onset, duration of epilepsy before FDG-PET, age when FDG-PET was performed, and PH or RH volume. The PH volume in patients with localized lesions on MRI (median 1,552 voxels; IQR 648–3,604 voxels) was significantly greater than that in patients without localized lesions (p = 0.03; median 627 voxels, IQR 250–1,617 voxels). No significant differences in RH volumes were found between patients with or without MRI-localized lesions; however, RH volume was significantly associated with the presence of SGTCS (p < 0.005).

The relationship between glucose hypometabolism and histology was examined. PH volume of neoplasm was found to show a trend toward larger volume (p = 0.053; neoplasm: median 2,109 voxels, IQR 1,201–4,809 voxels, gliosis: median 1,820 voxels, IQR 1,067–4,487 voxels, cortical dysplasia: median 1,189 voxels, IQR 322–2,755 voxels, vascular abnormality: median 590 voxels, IQR 250–7,104 voxels). No significant differences in RH volume were present with the different histology (p = 0.87).

Further analysis was performed that examined the relationship between glucose hypometabolism and the subtypes of cortical dysplasia. No statistically significant association between the type of cortical dysplasia and the extent of PH (p = 0.29) or RH (p = 0.96) was noted.

Distribution of RH

In frontal lobe epilepsy (n = 22), the most common site for RH was the temporal lobe (ipsilateral n = 6, contralateral n = 6) and contralateral frontal lobe (n = 7). Involvement of the ipsilateral parietal (n = 3), occipital (n = 1), basal ganglia (n = 2), thalamus (n = 1), and contralateral basal ganglia (n = 2), and thalamus (n = 1) were infrequent. In neocortical temporal lobe epilepsy (n = 24), contralateral temporal lobe (n = 14) hypometabolism was common. Other less common regions of RH included the frontal lobes (ipsilateral n = 5, contralateral n = 3), parietal lobes (ipsilateral n = 4, contralateral n = 2), basal ganglia (ipsilateral n = 7, contralateral n = 6) and thalamus (ipsilateral n = 3, contralateral n = 3). In parietal lobe epilepsy (n = 7), the distribution of remote hypometabolism was more scattered and involved the frontal lobes (ipsilateral n = 2, contralateral n = 2), ipsilateral temporal (n = 1) and occipital lobes (n = 1) and the contralateral parietal lobe (n = 2). In occipital lobe epilepsy (n = 15), RH was often present in the ipsilateral temporal lobe (n = 6) followed by the contralateral temporal lobe (n = 3). Rarely, RH occurred in the basal ganglia (ipsilateral n = 1, contralateral n = 1) and thalamus (ipsilateral n = 1, contralateral n = 1), ipsilateral frontal (n = 1) and parietal (n = 1) lobes, and contralateral occipital lobe (n = 1).

Discussion

In our cohort of patients with neocortical epilepsy, the identification of a structural lesion on MRI was a positive predictor of a good outcome. Conversely, a high RH volume on FDG-PET images had a negative prognostic association.

Thirty-four (79%) of our 43 patients with an MRI-identified lesion (mainly neoplasm and cortical dysplasia) became seizure-free. In patients with normal MRI, only 11 (44%) of the 25 patients were seizure-free. The presence of a structural lesion on MRI is an important factor for good outcome in our study and in studies at other centers (Cascino et al., 1992; Zentner et al., 1996; Kral et al., 2003a; Yun et al., 2006; Jeha et al., 2007; Elsharkawy et al., 2008). In two neocortical epilepsy series published recently, two thirds of the patients became seizure-free after surgery when a focal MRI lesion was present, whereas seizure freedom occurred in less than half of those without a definite MRI abnormality (Yun et al., 2006; Elsharkawy et al., 2008).

The majority of our patients with normal MRI showed cortical dysplasia on histologic examination. There were no truly cryptogenic cases, although one may consider the isolated finding of gliosis as a nonspecific pathology. In our cohort, we found no statistically significant association between histologic subtypes of cortical dysplasia and seizure outcome. In previously reported surgical series of patients with normal MRI, cortical dysplasia is a common pathologic substrate (Chapman et al., 2005; Lee et al., 2005; McGonigal et al., 2007). Some studies suggested that patients with type I cortical dysplasia had less favorable postsurgical seizure outcome when compared with those with type II dysplasia (Tassi et al., 2002; Kral et al., 2003b; Kim et al., 2009), whereas others found that cortical dysplasia histologic subtypes were not predictive of surgical outcome (Fauser et al., 2008; Krsek et al., 2009; Chang et al., 2011). At present, its relevance to surgical outcome is still unclear.

To date, our study is the only study examining the significance of remote glucose hypometabolism in neocortical epilepsy with multivariate analysis using a statistical map of glucose hypometabolism. The results indicated that a high RH volume predicted a poor outcome, independent of the presence of an MRI-identified structural abnormality. The majority of the patients without RH became seizure-free following surgery, whereas only one-third of those with widespread RH became seizure-free. Juhasz et al. (2001) studied 15 neocortical patients and on the basis of univariable analysis reported that the sizes of PH, RH, and total FDG-PET hypometabolism were not associated with surgical outcome (Juhasz et al., 2001). In this study, patients were highly selected and patients with single focal FDG-PET abnormalities that were consistent with EEG and MRI studies were excluded. Furthermore, glucose metabolism analysis was based on comparison of asymmetries that may conceal bilateral hypometabolism (Rubin et al., 1995). In another study, 86% (18 of 21 patients) of patients with normal or focal glucose hypometabolism achieved significantly better seizure outcome after surgery as compared to outcome improvement in only 30% (3 of 10 patients) of patients with widespread FDG-PET–identified abnormalities (Swartz et al., 1998). In temporal lobe epilepsy, studies have shown that the extent of hypometabolism outside of the epileptogenic temporal lobe is a predictor of outcome (Choi et al., 2003; Wong et al., 2010).

The reason for the significant association of RH volume with an unfavorable outcome is not evident from this study. Possible explanations could include the presence of multiregional epileptogenic zones, preexisting pathology, or seizure propagation beyond the epileptogenic zone with kindling of these regions. If RH reflects a diffuse or multiregional epileptogenic zone, then outcome failure could be related to residual pathology. For example, focal cortical dysplasia is one of the most frequent neuropathologic finding and occurs in 20–50% of patients undergoing epilepsy surgery (Tassi et al., 2002; Bast et al., 2006; Fauser et al., 2006; Luders & Schuele, 2006). However, the true prevalence is unknown and probably underestimated because of incomplete reporting (Sisodiya et al., 2009; Tassi et al., 2010). Focal cortical dysplasia can affect multiple lobes and is often not appreciated on MRI (Raymond et al., 1995; McGonigal et al., 2007; Fauser et al., 2009; Tassi et al., 2010; Blumcke et al., 2011). In such patients, some studies have suggested that a larger resection encompassing a greater proportion of metabolic or electrophysiologic abnormalities might achieve a better outcome (Bautista et al., 1999; Vinton et al., 2007; Kim et al., 2010). RH could be a marker for preexisting pathology capable of maturing into a new ictal focus and leading to subsequent surgical failure. The presence of a latent period following the initial epileptogenic injury before the emergence of spontaneous seizures is recognized, and that period may vary from months to more than 10 years (Walker et al., 2002). It is postulated that potential epileptogenic foci may mature at different rates and if subjected to another insult (a “second hit”) could accelerate the clinical expression of epilepsy (Engel, 1996; Walker et al., 2002; Yoon et al., 2003). This hypothesis is supported by a recent study demonstrating that a relatively mild insult on preexisting electrophysiologically silent cortical dysplasia can trigger epileptogenesis and the development of spontaneous seizures (Oghlakian et al., 2009). Future longitudinal studies examining the significance of RH in patients who relapse after years of initial seizure freedom will be provide more insight.

We postulate that RH cannot be attributed solely to widespread epileptogenicity and that its presence may be a consequence of seizure propagation. Our study results support this mechanism, by showing an association between the extent of RH and SGTCS, but not with histology, age of onset, or the presence of a MRI lesion. Although no study has reported that persistence of preoperative hypometabolism after surgery is associated with postoperative persistent seizures or with relapse after initial seizure freedom, some studies have shown that RH can improve following successful surgery, suggesting that metabolic disturbance may in part be a reversible effect of seizure propagation beyond the region of onset (Joo et al., 2005; Takaya et al., 2009). The role of seizure propagation into cortical areas beyond the lobe of onset in accounting for some proportion of RH has also been supported by studies correlating FDG-PET hypometabolism with electroclinical and ictal perfusion patterns (Bouilleret et al., 2002; Chassoux et al., 2004; Rusu et al., 2005).

There are some limitations in our study. Our sample size was insufficient for us to investigate the joint effect of the 12 individual brain regions in the multiple logistic regression analysis for each pathologic subtypes and the type of epilepsy, as the fitted model would have been overspecified. This retrospective study cannot provide the time between the last seizure and the 18F-FDG-PET, and this may affect the cerebral metabolic pattern (Savic et al., 1997; Gaillard et al., 2007). In recent studies, coregistration of FDG-PET and MRI was suggested to improve localization of the epileptogenic zone in patients with subtle or no MRI abnormality, and its use is encouraged as part of routine presurgical epilepsy workup (Salamon et al., 2008; Chassoux et al., 2010). Although we and others (Chapman et al., 2005; Jeha et al., 2007; Bien et al., 2009; Kim et al., 2009) did not find a significant association between a localizing FDG-PET and good surgical outcome, others had reported the contrary (Lee et al., 2005; Yun et al., 2006). It is possible our negative findings were due to inherent selection bias, as FDG-PET is part of our presurgical epilepsy evaluation and does inform the process of localization and the decision for surgery. Some series reported that the extent of surgical resection could influence seizure outcome (Bautista et al., 1999; Vinton et al., 2007; Kim et al., 2010). This was not examined in our study because postoperative volumetric MRI was not performed routinely in our patients. However, we did not find any statistically significant association between surgical outcome and volume of hypometabolic cluster involving the epileptogenic lesion and its adjacent contiguous regions. Lastly, regions of RH defined statistically using objective quantitative analysis were examined visually for MRI abnormalities. Although no abnormalities were detected visually, gray matter abnormalities detectable only with sophisticated analysis techniques such as voxel-based morphometry were not performed (Bonilha et al., 2006; Colliot et al., 2006).

In this study, patients with incomplete resection due to its proximity to eloquent cortex were excluded. In all the patients studied, the surgical resection was considered to be complete, as defined by complete removal of structural MRI lesion and regions with intracranial EEG onset. Despite this, not all patients were seizure-free. In these patients, we can only assume the resected site to be the epileptogenic zone for FDG-PET analysis. Nevertheless, other investigators have also found that complete resection of MRI or intracranial EEG abnormalities does not always guarantee complete seizure freedom (Zentner et al., 1997; Luyken et al., 2003; Kim et al., 2009; Krsek et al., 2009; Chang et al., 2010, 2011), and the findings of our FDG-PET abnormalities warrant further investigations.

In conclusion, accurate preoperative localization of the epileptogenic region is crucial to a successful outcome, and MRI plays an important role in such localization. In patients with widespread RH, epilepsy surgery may not result in complete seizure freedom despite complete removal of the MRI-located lesion. The findings in our study should be considered when determining the best approach to surgery (e.g., the role of invasive electrode evaluation), advice to patients, and postoperative follow-up and management.

Acknowledgment

This work was supported in part by University of Sydney Postgraduate Award, Millennium Institute Stipend, and Pfizer Neuroscience Research Grant to Dr Chong H. Wong.

Disclosure

Dr. Andrew Bleasel, Dr. Lingfeng Wen, Dr. Stefan Eberl, Dr. Karen Byth, Dr. Michael Fulham, Dr. Ernest Somerville, and Dr. Armin Mohamed report no conflicts of interest.

We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary