Predictors of seizure-free outcome after epilepsy surgery for pediatric tuberous sclerosis complex

Authors

  • Pavel Krsek,

    1. Department of Pediatric Neurology, 2nd Faculty of Medicine, Motol University Hospital, Charles University, Prague, Czech Republic
    Search for more papers by this author
  • Alena Jahodova,

    1. Department of Pediatric Neurology, 2nd Faculty of Medicine, Motol University Hospital, Charles University, Prague, Czech Republic
    Search for more papers by this author
  • Martin Kyncl,

    1. Department of Radiology, 2nd Faculty of Medicine, Motol University Hospital, Charles University, Prague, Czech Republic
    Search for more papers by this author
  • Martin Kudr,

    1. Department of Pediatric Neurology, 2nd Faculty of Medicine, Motol University Hospital, Charles University, Prague, Czech Republic
    Search for more papers by this author
  • Vladimir Komarek,

    1. Department of Pediatric Neurology, 2nd Faculty of Medicine, Motol University Hospital, Charles University, Prague, Czech Republic
    Search for more papers by this author
  • Petr Jezdik,

    1. Department of Measurement, Faculty of Electric, Czech Technical University Prague, Prague, Czech Republic
    Search for more papers by this author
  • Prasanna Jayakar,

    1. Department of Neurology and Comprehensive Epilepsy Program, Brain Institute, Miami Children's Hospital, Miami, Florida, U.S.A
    Search for more papers by this author
  • Ian Miller,

    1. Department of Neurology and Comprehensive Epilepsy Program, Brain Institute, Miami Children's Hospital, Miami, Florida, U.S.A
    Search for more papers by this author
  • Brandon Korman,

    1. Neuropsychology Section, Brain Institute and Behavioral Medicine, Miami Children's Hospital, Miami, Florida, U.S.A
    Search for more papers by this author
  • Gustavo Rey,

    1. Neuropsychology Section, Brain Institute and Behavioral Medicine, Miami Children's Hospital, Miami, Florida, U.S.A
    Search for more papers by this author
  • Trevor Resnick,

    1. Department of Neurology and Comprehensive Epilepsy Program, Brain Institute, Miami Children's Hospital, Miami, Florida, U.S.A
    2. Department of Neurology, University of Miami Miller School of Medicine, Miami, Florida, U.S.A
    Search for more papers by this author
  • Michael Duchowny

    Corresponding author
    1. Department of Neurology and Comprehensive Epilepsy Program, Brain Institute, Miami Children's Hospital, Miami, Florida, U.S.A
    2. Department of Neurology, University of Miami Miller School of Medicine, Miami, Florida, U.S.A
    • Address correspondence to Michael Duchowny, Department of Neurology, Brain Institute, Miami Children's Hospital, University of Miami Miller School of Medicine, 3200 S.W. 60th Court, Miami, FL, U.S.A. E-mail: michael.duchowny@mch.com

    Search for more papers by this author

Summary

Purpose

Variable predictors of postsurgical seizure outcome have been reported in children with tuberous sclerosis complex (TSC). We analyzed a large surgical series of pediatric TSC patients in order to identify prognostic factors crucial for selection of subjects for epilepsy surgery.

Methods

Thirty-three children with TSC who underwent excisional epilepsy surgery at Miami Children's Hospital were retrospectively reviewed. A total of 29 clinical, neuropsychological, electroencephalography (EEG), magnetic resonance imaging (MRI), and surgical variables were analyzed and related to seizure outcomes. Univariate Barnard's exact test, Wilcoxon's rank-sum test, and multivariate statistical Cox's model were used to examine the significance of associations between the variables and seizure outcome.

Key Findings

Eighteen patients (55%) have been seizure-free 2 years after (final) surgery; postoperative complications occurred in five subjects (15%). Complete removal of epileptogenic tissue detected by both MRI and intracranial EEG, regional scalp interictal EEG patterns, and agreement of interictal and ictal EEG localization were the most powerful predictors of seizure-free outcome. Other significant predictors included occurrence of regional scalp ictal EEG patterns, fewer brain regions affected by tubers, presence of preoperative hemiparesis, and one-stage surgery. Remaining factors such as age at seizure onset, incidence of infantile spasms or other seizure types, duration of epilepsy, seizure frequency, mental retardation, as well as types and extent of resections did not influence outcome.

Significance

Perioperative features rather than preoperative variables are the most important determinants of postsurgical seizure outcome in patients with TSC. Our findings may assist in the surgical management of these patients.

Epilepsy is a leading cause of morbidity in children with tuberous sclerosis complex (TSC), occurring in up to 92% of patients (Curatolo et al., 2005; Holmes et al., 2007). Because >50% of patients become medically refractory (Teutonico et al., 2008), epilepsy surgery is an increasingly important therapeutic option (Jansen et al., 2007; Madhavan et al., 2007; Bollo et al., 2008; Evans et al., 2012). Recent studies have shown that despite several specific challenges to epilepsy surgery planning in TSC populations, a significant proportion of patients (>60%) could become seizure-free with proper surgical candidate selection and evaluation (Lachhwani et al., 2005; Weiner et al., 2006; Moavero et al., 2010). Appropriate management of epilepsy surgery candidates based on identification of predictors of seizure-free/unfavorable seizure outcome is therefore of key importance.

One-stage surgery is safe and effective in TSC patients with a single seizure type, single or one large tuber, and convergent electrophysiologic data, (Koh et al., 2000; Romanelli et al., 2004; Hirfanoglu & Gupta, 2012). However, epilepsy surgery planning is complex in a majority of TSC patients owing to the presence of multiple cortical tubers rather than a single lesion, the frequent occurrence of multiple seizure types, and multifocal scalp electroencephalography (EEG) epileptiform activity.

Recent multicenter surveys reported younger age at seizure onset, history of infantile spasms, and multifocal interictal EEG findings (Madhavan et al., 2007), as well as the presence of tonic seizures and moderate or severe intellectual disability (Jansen et al., 2007) as significant predictors of poor postoperative outcome. Factors significantly related to seizure freedom were concordant and localized EEG and magnetic resonance imaging (MRI) findings (Lachhwani et al., 2005) and interictal (but not ictal) focality on scalp EEG (Madhavan et al., 2007). However, these results are not universally accepted. One of the meta-analyses, for example, found the residual tuber on magnetic resonance imaging (MRI) to be significantly associated with good seizure outcome, which is likely a coincidental finding (Madhavan et al., 2007).

To address these issues we analyzed clinical, EEG, MRI, neuropsychological, and surgical data in a large surgical cohort of children with TSC with respect to seizure outcome. The study aimed to identify prognostic factors crucial for selecting pediatric TSC patients for epilepsy surgery.

Methods

Patient selection and presurgical evaluation

Data from pediatric patients with TSC and medically refractory epilepsy who underwent excisional epilepsy surgery at the Miami Children's Hospital from 1996 to 2010 were reviewed retrospectively. We only included patients who had a definite diagnosis of TSC as defined in the revised clinical diagnostic criteria (Roach et al., 1998) and known seizure outcome at 2 years after (the last) surgery. Thirteen subjects were previously reported in published studies from Miami Children's Hospital (Koh et al., 1999; Koh et al., 2000).

Demographic and clinical variables were obtained in subjects. Patients underwent a comprehensive preoperative evaluation that always included long-term scalp video-EEG and high-resolution MRI. Selected patients also underwent 18F-fluorodeoxyglucose positron emission tomography (FDG-PET), ictal single photon emission computed tomography (SPECT), and functional MRI (fMRI); evaluation of these results was not included in the present study.

The influence of each of the following patient characteristics on seizure outcome was analyzed: family history of TSC and epilepsy, early psychomotor development delay (noted during the first year of life), age at seizure onset, incidence of individual seizure types including infantile spasms and status epilepticus, frequency of seizures, and details of neurologic examination.

Intellectual test results and/or overall adaptive functioning questionnaires available in 22 patients were reevaluated by two independent experts in the psychological assessment of children with neurocognitive disorders (BK and GR). Global functional ranking was determined because heterogeneous neuropsychological batteries had been utilized over the years in the assessment of the cases and also because some of the subjects could not be examined with common psychometric instruments owing to pervasive intellectual or developmental impairments. For the purpose of the study, patients were categorized into the following two groups: (1) Subjects with mental retardation (Mild to Severe Impairment), IQ ≤69, and (2) Subjects with Low Average Intelligence and Above, IQ >70.

All patients underwent preoperative 32-channel scalp video-EEG using the standard 10-20 system of electrodes with additional electrodes applied in selected cases. The following EEG features were correlated with postsurgical seizure outcomes: background activity slowing, types of interictal and ictal epileptiform abnormalities, and their concordance. Interictal epileptiform discharges and ictal EEG patterns were classified as either regional (appearing exclusively over a single lobe or in two contiguous regions such as centroparietal discharges) or nonregional. We also classified interictal and ictal epileptiform activity as originating from one hemisphere (regardless of extent) or independently from both hemispheres. Concordance of interictal and ictal EEG findings was defined as interictal and ictal epileptiform activity always localized to the same brain region (e.g., frontal, central, parietal, temporal, or occipital region(s) of the same hemisphere). Frontal lobe EEG concordance was further defined as colocalization to a defined region within the lobe.

All patients had at least one high-quality preoperative MRI scan performed with a 1.5-Tesla magnet including fluid-attenuated inversion recovery (FLAIR) sequences. MRI data were reevaluated independently by three experienced investigators (PK, MK, and MK) who were aware of the diagnosis but blinded to all other clinical, diagnostic, and seizure outcome data. For study purposes, the reviewers were required to localize cortical tubers to 11 anatomically defined cortical regions in both hemispheres (frontal central, mesial, convexity, polar, basal; temporal mesial and lateral; parietal mesial and lateral; occipital mesial and lateral). Tubers were defined as areas of cortical gray matter distortion with decreased or isointense subcortical signal intensity in T1-weighted images and increased signal intensity on T2-weighted and FLAIR sequences (Gallagher et al., 2010). In infants with unmyelinated white matter, tubers were characterized by hyperintense areas on T1-weighted images that were hypointense or isointense on T2-weighted images (Christophe et al., 1989). Discrepancies between the reviewers led to a case being re-reviewed together until a consensus was reached. If disagreement remained, the localization in question was omitted from the analysis.

Epilepsy surgery planning, surgical procedures, completeness of resection, and outcome

Surgical planning was determined at interdisciplinary case-conferences based on multimodal data (seizure semiology, neurologic status, interictal and ictal EEG, neuroimaging findings). In accord with the protocol of our center, decisions emphasized electrophysiologic investigations. In brief, when clinical/EEG and ictal SPECT data converged to one large tuber, or a specific region with several tubers, a one-stage resection guided by electrocorticography (ECoG) was undertaken. When noninvasive studies implied a broad, ill-defined region with multiple tubers, a two-stage procedure with subdural/intracerebral electrode implantation was considered necessary (Koh et al., 2000). Patients with a structurally ill-defined epileptogenic lesion, multifocal/not-lateralized scalp EEG findings, and nonlocalizing ictal SPECT were not referred for resective epilepsy surgery.

Seventeen patients underwent one-stage excisional procedures employing preexcision ECoG to define the epileptogenic zone and resection plane (localization of the below-defined significant ECoG abnormalities helped to specify resection planes in all the patients). Chronic invasive monitoring utilizing implanted subdural electrodes was performed in 16 patients. Only 2 of 33 patients were resected at more than one surgical site.

Surgical variables that were analyzed for relationship to seizure outcome included age at surgery, duration of epilepsy, side of surgery, type of the resection in relation to localization of cortical tubers (resection confined to a region of one tuber vs. area of more tubers), extent of resection (unilobar vs. multilobar), chronic invasive monitoring, reoperations, and completeness of the resection.

Completeness of the resection was determined by the epilepsy team composed of neurologists, neuroradiologists, and neurosurgeons at time of surgery, and was subsequently reviewed by the authors using primary data. In patients with TSC, the resection was considered complete only if a specific structurally abnormal region detected by MRI (usually one large tuber or a region with several tubers believed to represent the epileptogenic zone) and significant intracranial EEG abnormalities (defined below) were entirely removed. The criteria for evaluating intracranial EEG data have been published previously (Jayakar et al., 1994; Turkdogan et al., 2005; Krsek et al., 2009). In patients undergoing intraoperative electrocorticography, regions of active spiking with consistent focality, exhibiting rhythmic features such as trains of focal fast activity, or associated with focal attenuation of background, were considered significant and resected. Infrequent spikes and spikes without consistent focality were ignored. In chronically implanted subjects, the epileptogenic region was defined as a region exhibiting focal rhythmic activity, bursts of high frequency discharges, repetitive spiking, or electrodecremental patterns. Secondary foci that consistently activated during seizures were also included in the resection if they occurred in tissue adjacent or in regional proximity to the primary ictal focus. We defined secondary foci as cortical regions evidencing early spread of ictal activity and active independent spiking that were detected during intracranial recordings. Regions of frequent focal interictal spiking and background abnormalities with consistent focality were also considered significant. Slow waves occurring over widespread regions shortly after seizure onset were not regarded as critical for completeness of resection.

Postoperative seizure outcome 2 years after the final surgery was assessed during outpatient visits and telephone contact. Seizure outcome was classified according to Engel's classification scheme: Engel I, completely seizure-free, auras only or only atypical early postoperative seizures; Engel II, ≥90% seizure reduction or nocturnal seizures only; Engel III, ≥50% seizure reduction; and Engel IV, <50% seizure reduction. For the purpose of the study, subjects were classified as either “seizure-free” (Engel I group only) or “less favorable” (Engel II, III, and IV groups together).

Statistical analysis

Overall relationships between seizure outcomes and different categorical variables of interest were initially assessed using Barnard's exact test of significance of association within a contingency table format. The table columns equal to the outcome in two classes, that is, seizure-free and less favorable outcomes. The rows correspond to the partial tested variables. The tested null hypothesis was the independence of the categorical data within the contingency table.

The noncategorical variables (age at seizure onset, number of brain regions with tubers, age at surgery and duration of epilepsy) were tested using distribution-free Wilcoxon rank-sum test. The significance level of both univariate statistical tests was set to conventional 95%.

Multivariate statistical Cox's model of proportional hazards was then applied. The regression model assumes the variable “duration of epilepsy” is a random variable dependent on the explanatory variables. The univariate variables that were significant in the first tests were taken as the model explanatory. A linear hypothesis test of proportionality of the partial hazard functions using the estimated covariance matrixes in the Cox's model was utilized. The tests were unable to reject the null hypothesis, that is, the Cox model assumptions were satisfied.

Results

From an epilepsy surgery database of 594 patients (1996–2010), 38 subjects were diagnosed with TSC. Of these, five patients were lost to follow-up. A total of 33 patients (15 female and 18 male) were retained for analysis. Eighteen patients (55%) have been seizure-free 2 years after the (final) surgery. There were no subjects with persisting auras or early postoperative seizures; all patients were thus classified as Engel Ia. Seizure outcome in the remaining subjects was classified as follows: Engel II in five patients (15%), Engel III in four patients (12%), and Engel IV in six patients (18%). For study purposes, these 15 subjects were pooled into the “unfavorable” group.

Postoperative complications occurred in five cases and included subdural hematoma (n = 2), transient motor weakness (n = 2) (worsening of preexisting hemiparesis in one case), and diabetes insipidus in one child.

A total of 29 clinical, neuropsychological, EEG, MRI, and surgical variables were related to postsurgical seizure outcomes (Tables 1 and 2). Figure 1 represents statistical significances in 25 categorical variables. None of the four noncategorical variables was statistically significant at the chosen confidence level 99%.

Table 1. Relation of clinical data, neurological, and neuropsychological findings to seizure outcome.
CharacteristicSeizure-freeUnfavorablep-Value
  1. SGTCS, secondarily generalized tonic–clonic seizures.

  2. The values in parentheses are percentages. Significant findings (if p < 0.05 in Barnard's exact test) are in bold.

  3. a

    p-Value related to univariate analysis.

  4. b

    p-Value related to multivariate analysis.

1. Family history of TSC   
Positive (n = 7)4 (57)3 (43) 
Negative (n = 26)13 (50)13 (50) 
2. Family history of epilepsy   
Positive (n = 10)6 (60)4 (40) 
Negative (n = 23)11 (48)12 (52) 
3. Early psychomotor development retardation   
Yes (n = 17)10 (59)7 (41) 
No (n = 16)7 (44)9 (56) 
Age at seizure onset (years)   
All subjects: 2.56; 0.001–5 (mean; range)2.67; 0.001–52.5; 0.001–4 
4. Incidence of infantile spasms   
Yes (n = 12)7 (58)5 (42) 
No (n = 21)10 (48)11 (52) 
5. Incidence of status epilepticus   
Yes (n = 5)4 (80)1 (20) 
No (n = 28)13 (46)15 (54) 
6. Incidence of simple partial seizures   
Yes (n = 23)14 (61)9 (39) 
No (n = 10)3 (30)7 (70) 
7. Incidence of complex partial seizures   
Yes (n = 25)12 (48)13 (52) 
No (n = 8)5 (62.5)3 (37.5) 
8. Incidence of SGTCS   
Yes (n = 9)4 (44)5 (56) 
No (n = 24)13 (54)11 (46) 
9. Incidence of generalized seizures (e.g., atonic and myoclonic)   
Yes (n = 8)3 (37.5)5 (62.5) 
No (n = 25)14 (56)11 (44) 
10. Frequency of seizures   
Daily seizures (n = 29)15 (52)14 (48) 
Less frequent (n = 4)2 (50)2 (50) 
11. Neurologic finding   
Abnormal (n = 14)9 (64)5 (36) 
Normal (n = 19)8 (42)11 (58) 
12. Incidence of hemiparesis   
Yes (n = 9) 7 (78) 2 (22)  
No (n = 24) 10 (42) 14 (58) 0.045a, 0.41b
13. Incidence of nonlateralized neurologic deficits   
Yes (n = 10)6 (60)4 (40) 
No (n = 23)11 (48)12 (52) 
14. Incidence of mental retardation (in a subgroup of 22 subjects with available results of neuropsychological testing)   
Yes (n = 13)7 (54)6 (46) 
No (n = 9)3 (33)6 (67) 
Table 2. Relation of EEG features, MRI findings, and surgical variables to seizure outcome.
CharacteristicSeizure-freeUnfavorablep-Value
  1. The values in parentheses are percentages. Significant findings (if p < 0.05 in Barnard's exact test) are in bold.

  2. a

    p-Value related to univariate analysis.

  3. b

    p-Value related to multivariate analysis.

15. Background EEG activity   
Slow (n = 22)12 (55)10 (45) 
Normal (n = 11)5 (45)6 (55) 
Interictal epileptiform EEG activity   
16. Regional (n = 11) 9 (82) 2 (18)  
Nonregional (n = 22) 8 (36) 14 (64) 0.008a, 0. 0.010b
17. One-hemispheric (n = 20) 14 (70) 6 (30)  
Bilateral independent (n = 13) 3 (23) 10 (77) 0.005 a
Ictal EEG patterns   
18. Regional (n = 20) 13 (65) 7 (35)  
Nonregional (n = 13) 3 (23) 10 (77) 0.005a, 0.093b
19. One-hemispheric (n = 25)15 (60)10 (40) 
Bilateral independent (n = 8)2 (25)6 (75) 
20. Interictal and ictal EEG findings colocalized   
Yes (n = 12) 9 (75) 3 (25)  
No (n = 21) 8 (38) 13 (62) 0.023a, 0.004b
Number of brain regions with tubers   
All subjects: 7.9; 1–17 (mean; range) 6.4; 1–17 9.7; 6–17 0.023 a
Age at surgery (years)   
All subjects: 6.6; 0.1–17.6 (mean; range)5.4; 0.1–147.7; 0.3–17.6 
Duration of epilepsy (years)   
All subjects: 4.6; 0.1–17.3 (mean; range)3.6; 0.1–13.15.7; 0.3–17.3 
21. Invasive monitoring   
Yes (n = 16) 5 (31) 11 (69)  
No (n = 17) 12 (71) 5 (29) 0.018a, 0.241b
22. Side of surgery   
Right (n = 17)10 (59)7 (41) 
Left (n = 16)7 (44)9 (56) 
23. Type of surgery   
Around one tuber (n = 11)6 (55)5 (45) 
Area of more tubers (n = 22)11 (50)11 (50) 
24. Extent of surgery   
One lobe (n = 24)13 (54)11 (46) 
Multilobar (n = 9)4 (44)5 (56) 
25. Completeness of surgery   
Complete (n = 20) 15 (75) 5 (25)  
Incomplete (n = 13) 2 (15) 11 (85) <0.001a, 0.018b
Figure 1.

The p-Values of Barnard's exact test reflecting associations between individual categorical variables and seizure outcome. The red line shows the critical, 5% value. The features with bars below the line are statistically significant; the null hypothesis is rejected with 95% significance level. Variables depicted in bold were significant using multivariate statistical analysis.

Seizure outcome in relation to clinical and neuropsychological data

Family history of both epilepsy and TSC did not significantly influence seizure outcome. Significant pre-/perinatal risk factors such as hypoxia or bleeding as well as other risk factors for epilepsy such as head trauma and central nervous system (CNS) infection did not occur in our series. No association between early psychomotor development delay and seizure outcome was found.

Mean age at seizure onset was 1.05 years (range 1 day–5 years, median age was 0.44 year); 88% of patients had daily seizures. The association between age at seizure onset as well as seizure frequency and seizure outcome was not significant. Seizure outcome was not significantly influenced by the occurrence of several specific seizure presentations including infantile spasms, status epilepticus, simple and complex partial, and generalized and secondarily generalized tonic–clonic seizures.

Presence of abnormal presurgical neurologic findings and nonlateralized neurologic deficits did not influence seizure outcome. Seizure-free outcomes were noted, however, for patients with presurgical hemiparesis (Barnard's exact test, p = 0.045). All nine patients with hemiparesis had tubers in the central region. Five underwent resections including sensorimotor cortex (eight underwent one-stage surgeries; only one subject had a long-term intracranial EEG study). Despite the challenging epilepsy surgery management in these cases, only one child exhibited worsening of a preexisting hemiparesis and seven of nine patients became seizure-free. However, a contribution of presurgical hemiparesis to seizure outcome was not confirmed using multivariate analysis utilizing other variables that were significant on Barnard's exact test.

Subjects with and without mental retardation (Standard Score ranking below and above 70) did not differ in seizure outcome.

Seizure outcome in relation to EEG and MRI features

The presence of generalized background activity slowing did not influence seizure outcome. However, we did find a significant association between the presence of both ictal and interictal regional epileptiform abnormities on scalp EEG and seizure-free postsurgical outcome (Barnard's exact test, p = 0.005 and 0.008, respectively). The same association was identified when interictal epileptiform activity was classified as originating from one hemisphere or was bilaterally independent (Barnard's exact test, p = 0.005). When both ictal and interictal focality were entered into a multivariate regression model with other significant variables, contribution of interictal, but not ictal, regional epileptiform abnormalities to seizure-free outcomes was found to be significant (p = 0.02 and 0.093, respectively). Superior seizure outcomes were also found (by both analyses) in subjects with colocalized ictal and interictal epileptiform activity (Barnard's exact test, p = 0.023, Cox proportional multivariate analysis, p = 0.009). Analyses of MRI findings revealed that seizure-free outcomes significantly prevailed in patients who had fewer brain regions affected by tubers (Barnard's exact test, p = 0.023).

Seizure outcome in relation to surgical variables

Mean age at (first) surgery was 5.17 years (range 0.1–17.59 years; median age was 4.42 years) and mean duration of epilepsy was 4.56 years (range 0.1–17.34 years; median was 2.74 years). Both age at surgery and duration of epilepsy did not influence seizure outcome. Patients with one-stage surgeries had superior seizure outcomes compared to subjects undergoing long-term invasive monitoring (Barnard's exact test, p = 0.018). Nevertheless, one-stage surgery did not show a significant impact on seizure outcomes in the multivariate model (p = 0.24). Extent (one-lobe vs. larger resections) and side of surgery did not correlate with seizure outcome. We also found no differences in outcomes between patients with resections confined to a region of one tuber compared to those with surgeries in areas of more tubers. Four subjects underwent a reoperation (usually extending the previous resection); one of them was rendered seizure-free after a repeated surgery (statistical analysis was not performed because of the small number of reoperations). Finally, patients with complete resections had much better seizure outcomes compared to patients judged to have undergone incomplete removals of the epileptogenic zone. Both univariate and multivariate analyses confirmed a strong association between complete resection of the epileptogenic zone and seizure outcomes (Barnard's exact test, p < 0.001, Cox proportional multivariate analysis, p = 0.018).

Discussion

In our surgical series of pediatric TSC patients, the most powerful predictors of seizure-free outcome (proven by both univariate and multivariate analyses) included the following: (1) complete removal of the epileptogenic tissue as defined by both MRI and intracranial EEG, (2) occurrence of regional scalp interictal EEG epileptiform activity, and (3) agreement of interictal and ictal EEG localization. Other significant predictors (demonstrated by univariate tests) were (4) regional scalp ictal EEG patterns, (5) fewer brain regions affected by tubers, (6) presence of preoperative hemiparesis, and (7) one-stage surgery (without previous long-term invasive EEG). Several factors did not influence outcome including age at seizure onset, incidence of infantile spasms and other seizure types, duration of epilepsy, seizure frequency, mental retardation, and types and extent of resections.

The most important factor predicting postsurgical seizure freedom was complete removal of the epileptogenic tissue detected by both MRI (i.e., one large tuber or a region with several tubers supposed to represent the epileptogenic zone) and intracranial EEG findings. Seizure freedom was achieved in 75% of patients who had complete resections, but only in 15% of those with incomplete excisions of the epileptogenic region. Employing standardized criteria of completeness of resections assessment (Jayakar et al., 1994; Paolicchi et al., 2000; Krsek et al., 2009) thus predicted postsurgical seizure outcomes in TSC subjects, even more reliably than in our previous series of patients with focal cortical dysplasia (FCD): 70% of seizure-free subjects after complete and 22% after incomplete resections (Krsek et al., 2009).

Achieving complete resections in patients with TSC represents specific challenges due to the presence of multiple cortical tubers compared to the single lesions typically encountered in others surgical pathologies and the frequent occurrence of multifocal scalp EEG epileptiform activity. Diagnostic and surgical strategies vary tremendously between epilepsy centers (Jansen et al., 2007; Madhavan et al., 2007; Bollo et al., 2008). At Miami Children's Hospital, we employ a standardized algorithm of presurgical evaluation that emphasizes both electrophysiologic and neuroimaging investigations. In brief, when the clinical/EEG and ictal SPECT data converge to a specific region with one large tuber or several tubers, we perform a one-stage resection guided by ECoG. If the noninvasive studies suggest a broader ill-defined region with multiple tubers, we perform a two-stage procedure with subdural/intracerebral electrode implantation (Koh et al., 2000).

We wish to emphasize that the same algorithm has been used for our patients with FCD and resulted in exactly the same rate of postoperative seizure freedom: 55% (Krsek et al., 2009). We speculate that similar results in both series might be explained by the presence of common underlying structural pathology. It has been hypothesized that malformed tissue surrounding cortical tubers rather than the tubers themselves is the primary source of epileptic activity (Chandra et al., 2006; Bollo et al., 2008; Wong, 2008). Our approach in managing TSC cases is based on the clinically based presumption that perituberal dysplastic tissue might be identified preoperatively, and its surgical removal (in addition to resection of tubers) enhances the likelihood of achieving postsurgical seizure freedom.

Similar to FCD cases, most of the patients in our series (31 of 33 subjects) underwent resections at one site (only two patients had multiple tuberectomies). This stands in contrast to a multistaged technique that has been proposed for TSC patients with supposed multiple seizure foci (Weiner et al., 2004, 2006; Bollo et al., 2008). In this technique, multistaged and typically bilateral invasive intracranial monitoring is performed to identify both primary and secondary epileptogenic zones. After the surgical removal of the presumed primary seizure focus, initially unrecognized adjacent or distant epileptogenic zones are identified and operated on during a single hospitalization in order to reduce surgical failures attributed to these secondary sites. There are, however, no prominent differences in the outcome between our patients and patients who are multistaged (Weiner et al., 2006; Bollo et al., 2008).

It has been hypothesized that epilepsy in TSC might arise from a complex network in which a primary focus could generate multifocal seizures (Jacobs et al., 2008; Wong, 2008). We do not question the possibility of multiple epileptogenic zones in a subset of TSC patients, especially those with early onset catastrophic epilepsy (Weiner et al., 2006; Bollo et al., 2008). However, our experience suggests that although challenging, finding and removing a single primary focus will achieve success in a considerable proportion of TSC surgical candidates.

We found that children undergoing one-stage surgery guided by neuroimaging data and intraoperative ECoG experienced significantly better outcomes compared to subjects undergoing long-term intracranial EEG monitoring. This observation is open to bias, as long-term invasive EEG is indicated in complicated patients with presumably large, ill-defined, or multiple epileptogenic zones, or in subjects with overlapping epileptogenic and eloquent cortical areas. We nevertheless demonstrated that a significant proportion of TSC patients benefit from one-stage surgery employing ECoG and imaging-guidance. Several noninvasive diagnostic tests have been proposed to identify the epileptogenic zone in TSC, including ictal SPECT (Koh et al., 2000; Aboian et al., 2011), FDG-PET (Chandra et al., 2006; Wu et al., 2010), alpha-11C-methyl-L-tryptophan (AMT) PET (Kagawa et al., 2005), and magnetoencephalography/magnetic source imaging (MEG/MSI) (Wu et al., 2010). We believe that future progress in noninvasive diagnostic tests will increase the proportion of TSC patients managed without long-term intracranial EEG studies.

Seizure-free outcomes were significantly greater in patients with regional interictal and ictal scalp EEG epileptiform activity as well as in subjects with colocalized ictal and interictal EEG findings. Our results thus confirmed the importance of scalp EEG findings (especially regional interictal epileptiform activity proven to be significant predictor of postsurgical seizure-free outcome by both univariate and multivariate analyses) to select TSC children for successful epilepsy surgery. This observation has practical consequences for surgical planning in TSC, as scalp EEG data could be used to select potential surgical candidates from “structurally multifocal” subjects and thus improve surgical outcomes. An important observation with regard to this issue is that 90% of TSC patients evidence consistent region/regions of interictal epileptiform activity (Jansen et al., 2005). The same study reported that patients with one or two regions of epileptiform activity were older at seizure onset, often experienced complex partial seizures, had mild or no mental deficits, and might be better candidates for epilepsy surgery than subjects with multiple foci (Jansen et al., 2005). As several surprisingly good outcomes were reported for complex subjects (Romanelli et al., 2002; Weiner et al., 2006), we suggest that TSC patients with apparently multiple regions of epileptiform activity should be managed with specific diagnostic and treatment approaches such as the multistaged technique, although less favorable outcomes may currently be expected.

We noted better outcomes in patients with a lower number of involved brain regions on MRI. It seems logical that the epileptogenic zone might be more easily localized with fewer cortical tubers; however, overlapping of the epileptogenic zone with an eloquent cortical area is a more important predictor of success. In our experience, the crucial reason for surgical failure in TSC is overlapping of the epileptogenic zone with eloquent cortex rather than multifocal representation of the epileptogenic zone.

The presence of preoperative hemiparesis was the last significant variable associated with seizure-free postsurgical outcomes. The explanation might be that larger surgeries including eloquent cortical areas could more easily be offered to subjects with preexisting neurologic deficits than to those with a high risk of postoperative deterioration. Favorable outcome after extensive surgeries has been reported in pediatric epilepsy surgery series (Saneto & Wyllie, 2000). This observation should, however, be approached with caution as larger resections would logically be expected to have not only a higher likelihood of seizure freedom, but also a higher rate of complications. Moreover, the extent of surgery did not influence the incidence of seizure freedom in our series.

Finally, our study does not confirm the findings of recently published meta-analyses (Jansen et al., 2007; Madhavan et al., 2007). Variables related to seizure recurrence have been primarily studied, which included younger age at seizure onset, history of infantile spasms, multifocal interictal EEG findings (Madhavan et al., 2007), occurrence of tonic seizures, and mental retardation (Jansen et al., 2007). Except for the value of scalp EEG, outcome predictors derived from a review of the available literature and databases of different institutions appeared nonsignificant in our series. We therefore believe that infants with TSC, subjects with specific seizure types such as infantile spasms or bilateral seizures, as well as intellectually challenged patients should not be excluded from the epilepsy surgery consideration. The latest European clinical recommendations regarding epilepsy treatment in TSC suggested that epilepsy surgery is not indicated in patients with multiple seizure types and bilateral epileptic foci, but should be considered if there is a predominant seizure type that alters the quality of life (Curatolo et al., 2012).

We are aware of possible weaknesses and limitations of this study. This was a retrospective study and we only analyzed data from children who underwent surgery, not patients with TSC who were evaluated but eventually rejected for surgery or declined. These data are not available. Psychological testing was not available in all patients and neuropsychological batteries were heterogeneous due to a wide range of ages and intellectual capacities of our subjects. We also cannot exclude biases in the evaluation process related to the “philosophy” of our center.

Acknowledgment

Supported by grants IGA NT13357-4/2012 and by MH CZ – DRO, University Hospital Motol, Prague, Czech Republic 00064203.

Disclosure

None of the authors has any conflicts of interest to disclose. 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