Association of Seizure Spread With Surgical Failure in Epilepsy | Epilepsy and Seizures | JAMA Neurology | JAMA Network
[Skip to Navigation]
Sign In
Figure 1.  Engel Class I Seizure Freedom
Engel Class I Seizure Freedom

Seizure outcome after isolated anteromedial temporal resection. The cumulative probability of continuous seizure freedom is 83.9%, 77.8%, 70.9%, and 65.6% at 1, 2, 3, and 10 years or more after surgery, respectively.

Figure 2.  Predictors of Seizure Recurrence
Predictors of Seizure Recurrence

Analyses of Engel class I seizure freedom since surgery. A-C, Patients with a discordance of preoperative positron emission tomography (PET) with preoperative scalp electroencephalogram (EEG) (A) were significantly more likely to have seizure recurrence (hazard ratio [HR], 2.47; 95% CI, 1.03-5.94; P = .04), as were patients with a nonlesional pathology (B) (HR, 4.07; 95% CI, 1.31-12.68; P = .02) and those who were selected for intracranial EEG (iEEG) study (C) (HR, 6.52; 95% CI, 2.74-15.54; P < .001). P values from a comparison of the 3 predictive variables (A-C) were based on longitudinal Engel class outcome using a linear mixed-effects model and show iEEG selection retains statistical significance after controlling for the other 2 predictors.

Figure 3.  Rapid Spread
Rapid Spread

Rapid spread vs slow spread as represented by fast β power activity mapped onto 3-dimensional reconstructions of 4 different patients’ brains (A-D). The rapid spread examples (A and B) show a spread to the lateral temporal cortex, including the posterior temporal lobe and superior temporal gyrus, in less than 10 seconds. A quantification of fast β Power showed that patients whose seizures recurred after surgery (A and B) had statistically significant increases in β power in the lateral temporal cortex outside standard anteromedial temporal resection margins during the first 10 seconds of ictus (P < .05) compared with patients who were seizure-free after surgery (mean, 137.5%; SEM, 16.8% vs 93.4%; SEM, 4.6%). β Power differences between these 2 groups in the orbitofrontal and extratemporal regions did not reach statistical significance. The numbers between the upper and lower frames represent seconds associated with seizure onset at a time of 0 seconds. The small dots on the brain images represent the position of intracranial electroencephalogram (EEG) electrodes. The pale yellow coloration (100% baseline) represents no change in β power compared with the mean baseline power. The areas without blue dots and without coloration have no iEEG electrode and thus no available power data. Slow spread examples show an intense foci of β power increases that is limited to mesial temporal structures and the temporal pole almost entirely within the margins of anteromedial temporal resection.

aP < .05.

Figure 4.  Rapid Spread Outcomes
Rapid Spread Outcomes

Intracranial electroencephalogram (iEEG) analyses. A, Time to first spread and location of first spread. Patients whose seizures recurred after surgery predominately spread initially either within the anteromedial temporal lobe (AMT) or to the ipsilateral temporal neocortex (LaT) before spreading to other regions. Patients whose seizures recurred showed significantly shorter latency to the first spread on visual iEEG analyses (P < .05). B, Patients whose seizures spread from their site of onset in less than 10 seconds (11 [55%]) were significantly more likely (hazard ratio, 5.99; 95% CI, 1.7-21.1; P < .01) to have recurrent seizures after anteromedial temporal resection than patients selected for iEEG whose seizures did not spread in less than 10 seconds (9 [45%]). ExT indicates extratemporal cortex; OF, orbitofrontal cortex.

Table.  Patient Characteristicsa
Patient Characteristicsa
1.
Wiebe  S, Blume  WT, Girvin  JP, Eliasziw  M; Effectiveness and Efficiency of Surgery for Temporal Lobe Epilepsy Study Group.  A randomized, controlled trial of surgery for temporal-lobe epilepsy.  N Engl J Med. 2001;345(5):311-318. doi:10.1056/NEJM200108023450501PubMedGoogle ScholarCrossref
2.
Dwivedi  R, Ramanujam  B, Chandra  PS,  et al.  Surgery for drug-resistant epilepsy in children.  N Engl J Med. 2017;377(17):1639-1647. doi:10.1056/NEJMoa1615335PubMedGoogle ScholarCrossref
3.
Harroud  A, Bouthillier  A, Weil  AG, Nguyen  DK.  Temporal lobe epilepsy surgery failures: a review.  Epilepsy Res Treat. 2012;2012:201651.PubMedGoogle Scholar
4.
Ryvlin  P, Kahane  P.  The hidden causes of surgery-resistant temporal lobe epilepsy: extratemporal or temporal plus?  Curr Opin Neurol. 2005;18(2):125-127. doi:10.1097/01.wco.0000162852.22026.6fPubMedGoogle ScholarCrossref
5.
Barba  C, Rheims  S, Minotti  L,  et al.  Temporal plus epilepsy is a major determinant of temporal lobe surgery failures.  Brain. 2016;139(pt 2):444-451. doi:10.1093/brain/awv372PubMedGoogle ScholarCrossref
6.
Bartolomei  F, Chauvel  P, Wendling  F.  Epileptogenicity of brain structures in human temporal lobe epilepsy: a quantified study from intracerebral EEG.  Brain. 2008;131(pt 7):1818-1830. doi:10.1093/brain/awn111PubMedGoogle ScholarCrossref
7.
Spencer  SS, Berg  AT, Vickrey  BG,  et al; Multicenter Study of Epilepsy Surgery.  Predicting long-term seizure outcome after resective epilepsy surgery: the multicenter study.  Neurology. 2005;65(6):912-918. doi:10.1212/01.wnl.0000176055.45774.71PubMedGoogle ScholarCrossref
8.
West  S, Nolan  SJ, Cotton  J,  et al.  Surgery for epilepsy.  Cochrane Database Syst Rev. 2015;(7):CD010541.PubMedGoogle Scholar
9.
Semah  F, Picot  M-C, Adam  C,  et al.  Is the underlying cause of epilepsy a major prognostic factor for recurrence?  Neurology. 1998;51(5):1256-1262. doi:10.1212/WNL.51.5.1256PubMedGoogle ScholarCrossref
10.
Kang  JY, Wu  C, Tracy  J,  et al.  Laser interstitial thermal therapy for medically intractable mesial temporal lobe epilepsy.  Epilepsia. 2016;57(2):325-334. doi:10.1111/epi.13284PubMedGoogle ScholarCrossref
11.
Willie  JT, Laxpati  NG, Drane  DL,  et al.  Real-time magnetic resonance-guided stereotactic laser amygdalohippocampotomy for mesial temporal lobe epilepsy.  Neurosurgery. 2014;74(6):569-584. doi:10.1227/NEU.0000000000000343PubMedGoogle ScholarCrossref
12.
Cossu  M, Cardinale  F, Castana  L,  et al.  Stereoelectroencephalography in the presurgical evaluation of focal epilepsy: a retrospective analysis of 215 procedures.  Neurosurgery. 2005;57(4):706-718. doi:10.1227/01.NEU.0000176656.33523.1ePubMedGoogle ScholarCrossref
13.
Bourdillon  P, Isnard  J, Catenoix  H,  et al.  Stereo electroencephalography-guided radiofrequency thermocoagulation (SEEG-guided RF-TC) in drug-resistant focal epilepsy: results from a 10-year experience.  Epilepsia. 2017;58(1):85-93. doi:10.1111/epi.13616PubMedGoogle ScholarCrossref
14.
Morrell  MJ; RNS System in Epilepsy Study Group.  Responsive cortical stimulation for the treatment of medically intractable partial epilepsy.  Neurology. 2011;77(13):1295-1304. doi:10.1212/WNL.0b013e3182302056PubMedGoogle ScholarCrossref
15.
Sun  FT, Morrell  MJ, Wharen  RE  Jr.  Responsive cortical stimulation for the treatment of epilepsy.  Neurotherapeutics. 2008;5(1):68-74. doi:10.1016/j.nurt.2007.10.069PubMedGoogle ScholarCrossref
16.
Barbaro  NM, Quigg  M, Ward  MM,  et al.  Radiosurgery versus open surgery for mesial temporal lobe epilepsy: the randomized, controlled ROSE trial.  Epilepsia. 2018;59(6):1198-1207. doi:10.1111/epi.14045PubMedGoogle ScholarCrossref
17.
Ryvlin  P, Cross  JH, Rheims  S.  Epilepsy surgery in children and adults.  Lancet Neurol. 2014;13(11):1114-1126. doi:10.1016/S1474-4422(14)70156-5PubMedGoogle ScholarCrossref
18.
Bartolomei  F, Lagarde  S, Wendling  F,  et al.  Defining epileptogenic networks: contribution of SEEG and signal analysis.  Epilepsia. 2017;58(7):1131-1147. doi:10.1111/epi.13791PubMedGoogle ScholarCrossref
19.
Spencer  DD, Spencer  SS, Mattson  RH, Williamson  PD, Novelly  RA.  Access to the posterior medial temporal lobe structures in the surgical treatment of temporal lobe epilepsy.  Neurosurgery. 1984;15(5):667-671. doi:10.1227/00006123-198411000-00005PubMedGoogle ScholarCrossref
20.
Engel  J  Jr, Van Ness  PC, Rasmussen  TB, Ojemann  LM. Outcome with respect to epileptic seizures. In: Engel  J  Jr, ed.  Surgical Treatment of the Epilepsies. 2nd ed. New York, NY: Raven Press; 1993:609-621.
21.
Blumenfeld  H, Rivera  M, McNally  KA, Davis  K, Spencer  DD, Spencer  SS.  Ictal neocortical slowing in temporal lobe epilepsy.  Neurology. 2004;63(6):1015-1021. doi:10.1212/01.WNL.0000141086.91077.CDPubMedGoogle ScholarCrossref
22.
Englot  DJ, Yang  L, Hamid  H,  et al.  Impaired consciousness in temporal lobe seizures: role of cortical slow activity.  Brain. 2010;133(pt 12):3764-3777. doi:10.1093/brain/awq316PubMedGoogle ScholarCrossref
23.
Spencer  SS, Williamson  PD, Spencer  DD, Mattson  RH.  Human hippocampal seizure spread studied by depth and subdural recording: the hippocampal commissure.  Epilepsia. 1987;28(5):479-489. doi:10.1111/j.1528-1157.1987.tb03676.xPubMedGoogle ScholarCrossref
24.
Kim  JH, Guimaraes  PO, Shen  MY, Masukawa  LM, Spencer  DD.  Hippocampal neuronal density in temporal lobe epilepsy with and without gliomas.  Acta Neuropathol. 1990;80(1):41-45. doi:10.1007/BF00294220PubMedGoogle ScholarCrossref
25.
Spencer  DD, Spencer  SS.  Hippocampal resections and the use of human tissue in defining temporal lobe epilepsy syndromes.  Hippocampus. 1994;4(3):243-249. doi:10.1002/hipo.450040303PubMedGoogle ScholarCrossref
26.
Barba  C, Barbati  G, Minotti  L, Hoffmann  D, Kahane  P.  Ictal clinical and scalp-EEG findings differentiating temporal lobe epilepsies from temporal ‘plus’ epilepsies.  Brain. 2007;130(pt 7):1957-1967. doi:10.1093/brain/awm108PubMedGoogle ScholarCrossref
27.
Spencer  DD, Gerrard  JL, Zaveri  HP.  The roles of surgery and technology in understanding focal epilepsy and its comorbidities.  Lancet Neurol. 2018;17(4):373-382. doi:10.1016/S1474-4422(18)30031-0PubMedGoogle ScholarCrossref
28.
Dickten  H, Porz  S, Elger  CE, Lehnertz  K.  Weighted and directed interactions in evolving large-scale epileptic brain networks.  Sci Rep. 2016;6:34824. doi:10.1038/srep34824PubMedGoogle ScholarCrossref
29.
Epstein  CM, Adhikari  BM, Gross  R, Willie  J, Dhamala  M.  Application of high-frequency Granger causality to analysis of epileptic seizures and surgical decision making.  Epilepsia. 2014;55(12):2038-2047. doi:10.1111/epi.12831PubMedGoogle ScholarCrossref
30.
Friston  K, Moran  R, Seth  AK.  Analysing connectivity with Granger causality and dynamic causal modelling.  Curr Opin Neurobiol. 2013;23(2):172-178. doi:10.1016/j.conb.2012.11.010PubMedGoogle ScholarCrossref
31.
Jacobs  J, Zijlmans  M, Zelmann  R,  et al.  High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery.  Ann Neurol. 2010;67(2):209-220. doi:10.1002/ana.21847PubMedGoogle ScholarCrossref
32.
Tamilia  E, Park  EH, Percivati  S,  et al.  Surgical resection of ripple onset predicts outcome in pediatric epilepsy.  Ann Neurol. 2018;84(3):331-346. doi:10.1002/ana.25295PubMedGoogle ScholarCrossref
33.
Martinet  L-E, Ahmed  OJ, Lepage  KQ, Cash  SS, Kramer  MA.  Slow spatial recruitment of neocortex during secondarily generalized seizures and its relation to surgical outcome.  J Neurosci. 2015;35(25):9477-9490. doi:10.1523/JNEUROSCI.0049-15.2015PubMedGoogle ScholarCrossref
34.
Mesgarani  N, Cheung  C, Johnson  K, Chang  EF.  Phonetic feature encoding in human superior temporal gyrus.  Science. 2014;343(6174):1006-1010. doi:10.1126/science.1245994PubMedGoogle ScholarCrossref
Original Investigation
December 3, 2018

Association of Seizure Spread With Surgical Failure in Epilepsy

Author Affiliations
  • 1Department of Neurosurgery, Yale University School of Medicine, New Haven, Connecticut
  • 2Department of Neurology, Yale University School of Medicine, New Haven, Connecticut
  • 3Department of Neuroscience, Yale University School of Medicine, New Haven, Connecticut
JAMA Neurol. 2019;76(4):462-469. doi:10.1001/jamaneurol.2018.4316
Key Points

Question  Why do many patients continue to experience seizures after undergoing surgery for temporal lobe epilepsy?

Findings  In this cohort study of 118 patients, rapid seizure spread to regions outside resective medial temporal lobe margins was associated with seizure recurrence after surgery in patients who underwent identical medial temporal resections for drug-resistant temporal lobe epilepsy.

Meaning  In patients with drug-resistant epilepsy, areas of early seizure spread may represent an epileptogenic network and should be investigated as targets for resection or neuromodulation.

Abstract

Importance  Seizures recur in as many as half of patients who undergo surgery for drug-resistant temporal lobe epilepsy (TLE). Understanding why TLE is resistant to surgery in some patients may reveal insights into epileptogenic networks and direct new therapies to improve outcomes.

Objective  To characterize features of surgically refractory TLE.

Design, Setting, and Participants  Medical records from a comprehensive epilepsy center were retrospectively reviewed for 131 patients who received a standard anteromedial temporal resection by a single surgeon from January 1, 2000, to December 31, 2015. Thirteen patients were excluded for having less than 1 year of follow-up. Patients at the highest risk for seizure recurrence were identified. Intracranial electroencephalogram (iEEG) analyses generated 3-dimensional seizure spread representations and quantified rapid seizure spread. The final analyses of seizure outcome and follow-up data were performed in June 2017.

Main Outcomes and Measures  The Engel class seizure outcome following surgery was evaluated for all patients, defining seizure recurrence as Engel class II or greater. Intracranial recordings of neocortical grids/strips and depth electrodes were analyzed visually for seizure spread. Fast β power was projected onto reconstructions of patients’ brain magnetic resonance imaging scans to visualize spread patterns and was quantified to compare power within vs outside resective margins.

Results  Of 118 patients with 1 year of follow-up or more (mean [SD], 6.5 [4.6] years), 66 (55.9%) were women and 52 (44.1%) were men (median age, 39 years [range, 4-66 years]). The cumulative probability of continuous Engel class I seizure freedom since surgery at postoperative year 10 and afterward was 65.6%, with 92% of recurrences in years 1 to 3. Multivariable statistical analyses found that the selection for iEEG study was the most reliable predictor of seizure recurrence, with a mixed-effects model estimating that the Engel score in the iEEG cohort was higher by a mean (SD) of 1.1 (0.33) (P = .001). In patients with iEEG results, rapid seizure spread in less than 10 seconds was associated with recurrence (hazard ratio, 5.99; 95% CI, 1.7-21.1; P < .01). In the first 10 seconds of seizures, fast β power activity outside the resective margins in the lateral temporal cortex was significantly greater in patients whose seizures recurred compared with patients who were seizure-free (mean [SEM], 137.5% [16.8%] vs 93.4% [4.6%]; P < .05).

Conclusions and significance  Rapid seizure spread outside anteromedial temporal resection resective margins plays a significant role in the surgical failure of drug-resistant TLE. Seizure control after epilepsy surgery might be improved by investigating areas of early spread as candidates for resection or neuromodulation.

Introduction

Surgery for drug-resistant focal epilepsy, particularly temporal lobe epilepsy (TLE), has been shown to be superior to medical management, but 42% to 63% of patients go on to have seizures within 1 year after surgery.1,2 The reason surgery fails is not clear,3 but emerging evidence points to epileptogenic areas that are distinct from the seizure onset zone that are not targeted for resection.4-6 It is difficult to draw conclusions regarding the failures of epilepsy surgery from the existing literature that provides data from heterogenous interventions at multiple centers with multiple surgeons,5,7,8 sometimes without defining the anatomic resection.2,8

Temporal lobe epilepsy is the most common form of drug-resistant focal epilepsy that is treated with surgery.9 Evolving therapeutic and diagnostic techniques in epilepsy, such as stereoelectroencephalography (SEEG), stereotactic ablation,10-13 and responsive neurostimulation,14,15 have yet to improve seizure control over open resection.16 Identifying the epileptogenic foci or network that are responsible for continued or recurrent seizures postoperatively remains at the forefront of epilepsy research to improve seizure outcomes.17,18

In this study, we reviewed patients with medial TLE who underwent identical surgical resections by a single surgeon (D.D.S.). Some had intracranial monitoring with depth electrodes and extensive neocortical coverage confirming medial temporal lobe seizure onset. A cohort with long-term seizure outcome data was analyzed to identify the patients at highest risk for recurrent seizures after surgery.

Methods
Outcome Data

This study was conducted with institutional review board approval from the Yale Human Investigation Committee. Informed consent was obtained from patients prior to surgery. Records of patients who were undergoing the same anteromedial temporal resection (AMTR) without extratemporal resections by the senior author (D.D.S.) from January 1, 2000, until December 31, 2015, were collected from the Yale Epilepsy Center database. Medial temporal lobe responsibility for the seizure onset was determined primarily by magnetic resonance imaging (MRI) medial temporal pathology (mesial temporal sclerosis [MTS] or space-occupying lesion), concordant with scalp electroencephalogram (EEG) audiovisual monitoring. Patients were also evaluated with positron emission tomography (PET) scans and functional MRI for language and neuropsychological testing. The decision for AMTR was a consensus of the comprehensive epilepsy team at the surgical conference. Anteromedial temporal resection was a standard (right or left side) resection of the temporal pole (3-3.5 cm of the middle and inferior temporal gyri), a complete amygdalectomy, and the hippocampus divided where the tail turns behind the brainstem.19

At the epilepsy center, seizure onset dictated the resection site. For patients for whom intracranial EEG (iEEG) was undertaken, seizure onset was defined as the first electrode to show ictal activity. If the seizure onset was within the medial temporal lobe (amygdala, hippocampus, and entorhinal cortex), then AMTR was performed. For patients with a simultaneous onset in areas outside the medial temporal lobe, extended lateral temporal resections or extratemporal resections (most frequently frontal cortex) were sometimes undertaken depending on clinical and anatomical variables. Such cases of extended resections or extratemporal resections are not included in this study.

Of the 131 patients who met surgical criteria, 118 (90.1%) had 1 year or more of follow-up. Patients without 1-year follow-up were excluded from analyses. Engel class seizure outcomes20 were assigned with freedom from disabling seizures as a class I outcome (Figure 1 and Figure 2). Pathology results were determined by a postoperative pathology report. Nonlesional pathology results were usually nonspecific gliosis. Magnetic resonance imaging, PET, and scalp EEG were categorized by experts at the surgical conference as either localizing to the medial temporal lobe, multifocal lateralization to a hemisphere (without lobe localization), bilateral involvement (no lateralization), or no abnormality. Positron emission tomography scans were not available for 33 patients (28.0%).

Intracranial EEG was a combination of neocortical grids and strips and depth electrodes sampling medial structures. For an expanded explanation of seizure-spread analyses, see eMethods in the Supplement. Raw iEEGs were read by an epileptologist (P.F.) who was masked to patient outcomes. Seizure spread was defined as an ictal pattern at electrode contacts that was at least 2 cm from the seizure onset region. Thirty patients (25.4%) underwent iEEG, of whom 20 (66.7%) had traces available for review20 and were included in the time-to-first-spread analysis and Kaplan-Meier analysis of rapid spread (11 [55%]) vs slow spread (9 [45%]).

Magnetic resonance imaging was performed after intracranial electrode implantation, and 3-dimensional reconstructions of patients’ brains were constructed with power data that were mapped onto positions of intracranial electrodes. To analyze the seizure spread outside the surgical resective margins in Figure 3 and Figure 4A, brains were segmented into AMTR resection area19 (AMT), ipsilateral lateral temporal lobe outside resective margins (LaT), ipsilateral orbitofrontal cortex (OF), and extratemporal (excluding orbitofrontal) areas (ExT).21 Three-dimensional reconstruction data were missing for 4 of the 20 patients (20%) and these were excluded from the power analysis in Figure 3 (total, 16; recurrent seizures, 7 [43.8%]; seizure-free, 9 [56.2%]).

A fast Fourier transform analysis was performed on EEG signals. One-second signal segments and signal power were calculated for each electrode for δ (0.4 to <4 Hz), θ (4 to <8 Hz), α (8-13 Hz), β (>13 to <25 Hz), and γ (25-50 Hz) frequency bands.22 The average ictal percentage of β power of each electrode was averaged within each anatomical segment for each patient, with values for patients with recurring seizures vs patients without seizures compared using unpaired t tests with threshold for significance of α = .05.

Kaplan-Meier curves were generated to analyze continuous Engel class I seizure freedom since surgery (Figures 1, 2, and 4), with Kaplan-Meier P values calculated by log-rank (Cox proportional hazards assumption), Mantel-Haenszel hazard ratios (HRs), and 95% confidence intervals using GraphPad Prism, version 7.0a (GraphPad). A linear mixed-effects model was used with R software (version 1.1.442; R Foundation) using the lme4 extension package to analyze the association of multiple predictors of outcomes that were found to be significant on the Kaplan-Meier analysis (Figure 2D). Time-to-first-spread (Figure 4A) was compared in patients with recurring seizures vs patients without seizures using a 2-tailed Mann-Whitney U test (GraphPad). For the quantitative analysis of seizure spread (Figure 3), the ictal percentage of β power of each electrode was averaged within each anatomical segment to create 1 data point per anatomical segment for each patient, with recurred vs seizure-free groups compared using 2-tailed, unpaired t tests (GraphPad). The threshold for significance for all of these tests was α = .05.

Results

As shown in the Table, there were 118 patients older than 16 years who met the inclusion criteria of isolated AMTR with 1 or more years of follow-up. The mean (SD) follow-up of this cohort was 6.5 (4.6) years (median, 6 years [range, 1-17 years]), and 42 (35.6%) had 10 or more years of follow-up. Seizure outcomes are described in Figure 1, with continuous Engel class I seizure freedom analyzed using Kaplan-Meier analyses. Most seizure recurrences (33 of 36 [91.7%]) occurred during the first 3 postoperative years. Figure 1 shows a 65.6% cumulative probability of continuous Engel class I seizure freedom at 10 years and afterward. By year 10, of all the patients who had Engel class I seizures, 15 of 31 (48.4%) did not have any more seizures (Engel IA: no seizures and no auras since surgery), with the other 16 (51.6%) divided among those with auras (Engel IB, 4 [12.9%]), those with some seizures after surgery that were subsequently controlled for 2 years or longer (Engel IC, 6 [19.4%]), and those with seizures during antiepileptic drug withdrawal trials (Engel ID, 6 [19.4%]).

Of the variables that were analyzed as predictors of seizure recurrence (dropping out of Engel class I), the 3 that were found to be statistically significant are described in Figure 2: (1) selection for iEEG, (2) nonlesional pathology, and (3) discordance of preoperative PET with scalp EEG. When analyzed using a multivariable mixed-effects statistical model to predict worse (higher) Engel class outcomes, the selection for iEEG remained statistically significant (P = .001) when controlling for the other variables. While there were 30 patients (25.4%) who were selected for iEEG of the total patient cohort, more than half of these patients experienced recurrent seizures, having an HR of 6.5 (95% CI, 2.7-15.5) compared with patients who were not selected for iEEG.

Because of the high incidence of seizure recurrence in the iEEG subgroup, iEEG data were scrutinized for electrophysiological patterns to explain why surgeries failed. When the iEEG traces were analyzed for time to first seizure spread after onset, the range was 1 to 55 seconds (mean [SD], 12.25 [12.90] seconds; median, 8 seconds [interquartile range, 10.25]). Nine of 10 patients (90%) who had recurrent seizures had their first seizure spread less than 10 seconds after onset. In this group, all began medially before rapidly spreading outside the resective margins.

Examples of this rapid-spread phenotype are shown in Figure 3A and B, with fast β activity—previously shown to be associated with low-voltage fast seizure activity22—projected onto 3-dimensional reconstructions of the patients’ brains. This rapid spreading is contrasted with the patients in Figure 3C and D who did not show a rapid seizure spread electrographically and in whom the β activity was confined to medial structures for at least 10 seconds after seizure onset. Patients whose seizures recurred showed significant increases in fast β power in the LaT outside standard AMTR margins compared with patients who did not have seizures after undergoing surgery. While the mean β power in areas other than the lateral temporal lobe trended higher in some patients who had recurring seizures, our limited sample size did not detect significant differences when pooling data in these other regions. The differences in β power within AMTR margins for those with recurring seizures (mean [SEM], 184.8% [38.4%]) vs those without seizures (mean [SEM], 153.2% [24.5%]) were not significant (P = .48). The differences in β power in the OF between those with recurring seizures (mean [SEM], 125.6% [17.9%]) and those without seizures (mean [SEM], 98.8% [6.7%]) were not significant (P = .15). The differences in β power in the ExT between those with recurring seizures (mean [SEM], 120.8% [8.3%]) and those without seizures (mean [SEM], 103.0% [6.4%]) were not significant (P = .10).

Most patients with a rapid spread first showed seizure activity (Figure 4A) that spread either within the AMT or to the LaT before seizure activity was detected in the OF or ExT on visual iEEG analyses (Figure 4A). This pattern fits with previously described patterns of seizure spread in TLE.22,23 In addition, patients in the iEEG cohort with seizure spread in less than 10 seconds were significantly more likely to have recurrent seizures than their counterparts who did not show a rapid spread (Figure 4B). These data provide quantitative evidence that rapidly spreading seizures and the early ictal involvement of cortical structures outside resective margins underlie a substantial portion of AMTR failures and have implications for guiding therapeutic interventions.

Discussion

In this study, we described a large cohort of patients who underwent standard AMTR for medial TLE by a single surgeon with long-term follow-up. A limitation of previous studies has been the control for surgical intervention (multiple centers, multiple surgeons, disconnection, and tailored surgeries), which is addressed here using a single-surgeon single-center consecutive series of patients who received identical operations. This study described the rapid spread of medial temporal lobe seizure onset to the LaT, posterior to resective margins, as a feature that underlies a substantial portion of surgical failures. In addition, this study visualized and quantified these rapidly spreading seizures.

The results are best explained by attributing epileptogenic potential to sites of early seizure spread that were not included in resection. This mechanism of failure implies that a distributed epileptogenic network rather than a single epileptogenic focus may underlie surgically refractory epilepsy. A shift toward treating the seizure network rather than a seizure-onset focus could alter the surgical and neuromodulatory management of focal epilepsy as well as guide electrode placement for patients who require iEEG study.

In TLE that is resistant to surgery, histopathologic and electrographic studies have suggested an epileptogenic involvement of cortical areas outside medial temporal structures, even when hippocampal onset is observed. At the epilepsy center where this study was undertaken, volumetrically normal hippocampi are an indication for iEEG study even when scalp EEG and other noninvasive evaluations lateralize to 1 temporal lobe. In past analyses of such cases in which the ictal onset localized to 1 hippocampus and an AMTR was performed, the hippocampal neuronal loss was milder than typical of MTS (eg, 25% neuronal loss rather than >50% loss of MTS24) and was associated with lower rates of seizure control.25 Because the neuronal loss resembled hippocampi that were resected adjacent to developmental tumors in these cases (termed paradoxical medial temporal lobe epilepsy), it was speculated that the seizure source may originate in a path that was projecting to that hippocampus and not within it.25

More recent studies on surgical failures in TLE have pointed to a subset of patients who have a primary seizure onset in the temporal lobe with an “epileptogenic zone”5,26 that extends to nearby structures outside the standard resective margins, termed temporal-plus epilepsy (TPE).5,26 Similarities in seizure recurrence rates, as well as an overlap in areas of neocortical involvement, between the rapid-spread cohort described in this study and TPE subgroups5 suggest that these analyses of failed medial TLE surgery are converging.

The use of SEEG (ie, depth electrodes) without subdural strips and grids was the primary method for describing TPE. However, it is unclear whether SEEG has the ability to cover the distributed sites of neocortical spread that are detailed in this study. Seizure activity can only be recorded from areas of electrode coverage; therefore, analyses of seizure networks will be biased and limited by the extent of electrode coverage. While a detailed electrographic description of TPE was not provided in prior reports,4,5,26 the emphasis appears to be on seizure onset with regard to the epileptogenic zone. In this study, because of the coverage from subdural grid and strip electrodes that were used in combination with depth electrodes, the neocortical spread was broadly visualized. Still, our coverage may have been insufficient to fully characterize extratemporal involvement in such limbic regions as the cingulate and insula. These data point to an early spread rather than simultaneous onset as the marker of the poorest outcome.

An epileptogenic lateral temporal cortex in TLE has been implicated elsewhere in the literature, including that on TPE.5,26 Notably, SEEG studies aimed at defining epileptogenic zones have pointed to lateral temporal structures and areas of early seizure involvement as important epileptogenic nodes6 in patients with worse seizure control after surgery, which agrees with this study’s results. Moreover, if seizure networks that extend into the lateral temporal cortex do underlie a portion of surgically refractory TLE, one would predict that ablative techniques that are limited to medial temporal structures would show lower rates of seizure freedom than an open resection, which includes portions of the lateral cortex. Thus far, the literature indicates this.10,11,13,16

Ictal onset, rather than the seizure spread that is described, was the major factor taken into account in planning resections. To our knowledge, there is no current consensus on how early is too early in a seizure’s progression to suggest that an area of spread might generate seizures after the removal of the onset zone. The term rapid is used to describe the seizure spread detailed in this study, but prior to this analysis, extramedial temporal epileptogenesis requiring an extended or extratemporal resection at this center was typically only considered for simultaneous onsets that were outside AMTR margins. These data support expanding the epileptogenic zone to a multinodal network definition.

While this study’s data point to a rapid spread to regions outside the medial temporal lobe, this study may be statistically underpowered to detect clinically significant seizure spread in areas other than the lateral temporal neocortex (eg, orbitofrontal and extratemporal regions). Onset and early involvement in seizure activity have been associated with epileptogenicity,6 but to our knowledge, a discrete timeline of how long after onset is early enough to warrant the therapeutic targeting of such an area has yet to be delineated. This study provides evidence that rapid seizure propagation to the cortex beyond resection margins is associated with seizure recurrence.

Strengths and Limitations

This study had several limitations. First, this is a retrospective cohort and thus findings should be interpreted with appropriate caution until they are validated in a prospective manner. In addition, while the long-term follow-up on many patients is a strength in this study, a disadvantage is that system changes in medical centers and evolving methods for data collection resulted in incomplete data sets for some patients. Elements of raw iEEG data were not readily available for review to determine seizure spread in 10 of the 30 patients (33.3%) who underwent intracranial monitoring in our cohort: 5 (50%) with seizure recurrence and 5 (50%) who did not have seizures after surgery. Similarly, the generation of 3-dimensional reconstructions and anatomical segmentation for the power analysis was based on a confluence of patient data from several different sources, which were incomplete for 4 patients (3.9%), including 3 who showed a rapid spread and 1 with a slow spread, which possibly led to selection bias. Despite these drawbacks, the findings based on iEEG analyses are based on sample sizes that are comparable with related work in the literature.6,22

Regarding implications for the surgical management of drug-resistant TLE, this study’s data support a potential network theory of epileptogenicity in which nodes outside the seizure onset zone are implicated in seizure generation.18,27,28 The extratemporal cortex may provide sites of rapid spread that were not sampled. Our combined depth and neocortical studies most frequently provided a coverage of the lateral temporal, parietal, and frontal lobes with the depth electrode sampling of the orbitofrontal and medial frontal lobes. However, most patients did not have extensive coverage of the cingulate gyrus or the insular cortex, which may be suspected to mimic medial temporal lobe scalp EEG and semiology. These extratemporal limbic nodes could be critical to the epileptogenic network onset and going forward should be sampled when an iEEG study of presumed medial temporal onset is performed. The present approach to these patients takes advantage of the robotic depth accuracy together with our experience in neocortical surface coverage.

To address this problem therapeutically, extending the resection to include areas to which the seizures rapidly spread is one possible route when the spread pattern includes dispensable cortical areas, such as the nondominant lateral temporal cortex. When an extended resection is not possible (eg, in the dominant temporal lobe, insula, or frontal neocortex), one might place responsive neurostimulation electrodes at these sites. Responsive neurostimulation electrodes would allow long-term recordings to validate sites that are responsible for recurrent seizures as well as serve to test whether these sites are responsive to neuromodulation.27 When extratemporal sites, such as the parietal lobe or insula (either dominant or nondominant), are identified as a rapid-spread network, resection, ablation, or neurostimulation must be carefully weighed with regard to a possible adverse outcome. While these patients may likely fail an AMTR, the effectiveness of possible alternative approaches is not yet clear. Both extended resection and responsive neurostimulation will demand medial depth electrode and lateral temporal coverage for planning and a clear definition of the projected network nodes.

Regarding other future directions, the focus of this study was ictal data, but there may be clues in interictal connectivity29,30 changes and an analysis of high-frequency oscillations31,32 that could further distinguish epileptogenicity using iEEG data. The intracranial electrode sampling rate was less than 2000 Hz for the years of this data analysis, but the present system will allow for the addition of high-frequency oscillation analysis in patients who show a rapid spread.

Conclusions

The concept of epileptogenesis as a network phenomenon has been gaining more traction over the past few years and computational methods to support this are beginning to emerge.18,27,33 However, to our knowledge, it has yet been proven which nodes in a pathway that projects from a demonstrated ictal focus are epileptogenic and capable of generating seizures. The limiting factor for understanding neural networks that underlie epileptogenicity may be the quality and spatial coverage of intracranial data. Magnifying the granularity of iEEG data (eg, with high-density grids) promises to elucidate the normal physiology of the indispensable cortex.34 Expansions of technology for both diagnostic and research purposes may more clearly elucidate the mechanisms of failure in epilepsy surgery.27 Such progress, combined with improving electrophysiologic analyses that target epileptic networks, could allow patients who are currently deemed untreatable to have the possibility of seizure freedom.

Back to top
Article Information

Corresponding Author: Dennis D. Spencer, MD, Department of Neurosurgery, Yale University, 333 Cedar St, New Haven, CT 06520-8018 (dennis.spencer@yale.edu).

Accepted for Publication: November 1, 2018.

Published Online: December 3, 2018. doi:10.1001/jamaneurol.2018.4316

Author Contributions: Dr Andrews had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Andrews, Gummadavelli, Farooque, Bonito, Blumenfeld, Spencer.

Acquisition, analysis, or interpretation of data: Andrews, Gummadavelli, Farooque, Bonito, Arencibia, Spencer.

Drafting of the manuscript: Andrews, Gummadavelli, Bonito, Blumenfeld.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Andrews, Gummadavelli, Arencibia, Blumenfeld.

Obtained funding: Andrews, Spencer.

Administrative, technical, or material support: Andrews, Gummadavelli, Bonito, Arencibia, Spencer.

Supervision: Andrews, Farooque, Blumenfeld, Spencer.

Other - inntracranial EEG visual analysis observation of trend: Farooque.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported in part by the Swebilius Foundation (Dr Spencer), Howard Hughes Medical Institute (Dr Andrews), and Citizens United for Research in Epilepsy (Dr Andrews).

Role of the Funder/Sponsor: The funding sources did not have any role in design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank Jason Gerrard, MD, PhD, Yale University, for his helpful comments on the data analysis throughout the study. He was not compensated for his contribution.

References
1.
Wiebe  S, Blume  WT, Girvin  JP, Eliasziw  M; Effectiveness and Efficiency of Surgery for Temporal Lobe Epilepsy Study Group.  A randomized, controlled trial of surgery for temporal-lobe epilepsy.  N Engl J Med. 2001;345(5):311-318. doi:10.1056/NEJM200108023450501PubMedGoogle ScholarCrossref
2.
Dwivedi  R, Ramanujam  B, Chandra  PS,  et al.  Surgery for drug-resistant epilepsy in children.  N Engl J Med. 2017;377(17):1639-1647. doi:10.1056/NEJMoa1615335PubMedGoogle ScholarCrossref
3.
Harroud  A, Bouthillier  A, Weil  AG, Nguyen  DK.  Temporal lobe epilepsy surgery failures: a review.  Epilepsy Res Treat. 2012;2012:201651.PubMedGoogle Scholar
4.
Ryvlin  P, Kahane  P.  The hidden causes of surgery-resistant temporal lobe epilepsy: extratemporal or temporal plus?  Curr Opin Neurol. 2005;18(2):125-127. doi:10.1097/01.wco.0000162852.22026.6fPubMedGoogle ScholarCrossref
5.
Barba  C, Rheims  S, Minotti  L,  et al.  Temporal plus epilepsy is a major determinant of temporal lobe surgery failures.  Brain. 2016;139(pt 2):444-451. doi:10.1093/brain/awv372PubMedGoogle ScholarCrossref
6.
Bartolomei  F, Chauvel  P, Wendling  F.  Epileptogenicity of brain structures in human temporal lobe epilepsy: a quantified study from intracerebral EEG.  Brain. 2008;131(pt 7):1818-1830. doi:10.1093/brain/awn111PubMedGoogle ScholarCrossref
7.
Spencer  SS, Berg  AT, Vickrey  BG,  et al; Multicenter Study of Epilepsy Surgery.  Predicting long-term seizure outcome after resective epilepsy surgery: the multicenter study.  Neurology. 2005;65(6):912-918. doi:10.1212/01.wnl.0000176055.45774.71PubMedGoogle ScholarCrossref
8.
West  S, Nolan  SJ, Cotton  J,  et al.  Surgery for epilepsy.  Cochrane Database Syst Rev. 2015;(7):CD010541.PubMedGoogle Scholar
9.
Semah  F, Picot  M-C, Adam  C,  et al.  Is the underlying cause of epilepsy a major prognostic factor for recurrence?  Neurology. 1998;51(5):1256-1262. doi:10.1212/WNL.51.5.1256PubMedGoogle ScholarCrossref
10.
Kang  JY, Wu  C, Tracy  J,  et al.  Laser interstitial thermal therapy for medically intractable mesial temporal lobe epilepsy.  Epilepsia. 2016;57(2):325-334. doi:10.1111/epi.13284PubMedGoogle ScholarCrossref
11.
Willie  JT, Laxpati  NG, Drane  DL,  et al.  Real-time magnetic resonance-guided stereotactic laser amygdalohippocampotomy for mesial temporal lobe epilepsy.  Neurosurgery. 2014;74(6):569-584. doi:10.1227/NEU.0000000000000343PubMedGoogle ScholarCrossref
12.
Cossu  M, Cardinale  F, Castana  L,  et al.  Stereoelectroencephalography in the presurgical evaluation of focal epilepsy: a retrospective analysis of 215 procedures.  Neurosurgery. 2005;57(4):706-718. doi:10.1227/01.NEU.0000176656.33523.1ePubMedGoogle ScholarCrossref
13.
Bourdillon  P, Isnard  J, Catenoix  H,  et al.  Stereo electroencephalography-guided radiofrequency thermocoagulation (SEEG-guided RF-TC) in drug-resistant focal epilepsy: results from a 10-year experience.  Epilepsia. 2017;58(1):85-93. doi:10.1111/epi.13616PubMedGoogle ScholarCrossref
14.
Morrell  MJ; RNS System in Epilepsy Study Group.  Responsive cortical stimulation for the treatment of medically intractable partial epilepsy.  Neurology. 2011;77(13):1295-1304. doi:10.1212/WNL.0b013e3182302056PubMedGoogle ScholarCrossref
15.
Sun  FT, Morrell  MJ, Wharen  RE  Jr.  Responsive cortical stimulation for the treatment of epilepsy.  Neurotherapeutics. 2008;5(1):68-74. doi:10.1016/j.nurt.2007.10.069PubMedGoogle ScholarCrossref
16.
Barbaro  NM, Quigg  M, Ward  MM,  et al.  Radiosurgery versus open surgery for mesial temporal lobe epilepsy: the randomized, controlled ROSE trial.  Epilepsia. 2018;59(6):1198-1207. doi:10.1111/epi.14045PubMedGoogle ScholarCrossref
17.
Ryvlin  P, Cross  JH, Rheims  S.  Epilepsy surgery in children and adults.  Lancet Neurol. 2014;13(11):1114-1126. doi:10.1016/S1474-4422(14)70156-5PubMedGoogle ScholarCrossref
18.
Bartolomei  F, Lagarde  S, Wendling  F,  et al.  Defining epileptogenic networks: contribution of SEEG and signal analysis.  Epilepsia. 2017;58(7):1131-1147. doi:10.1111/epi.13791PubMedGoogle ScholarCrossref
19.
Spencer  DD, Spencer  SS, Mattson  RH, Williamson  PD, Novelly  RA.  Access to the posterior medial temporal lobe structures in the surgical treatment of temporal lobe epilepsy.  Neurosurgery. 1984;15(5):667-671. doi:10.1227/00006123-198411000-00005PubMedGoogle ScholarCrossref
20.
Engel  J  Jr, Van Ness  PC, Rasmussen  TB, Ojemann  LM. Outcome with respect to epileptic seizures. In: Engel  J  Jr, ed.  Surgical Treatment of the Epilepsies. 2nd ed. New York, NY: Raven Press; 1993:609-621.
21.
Blumenfeld  H, Rivera  M, McNally  KA, Davis  K, Spencer  DD, Spencer  SS.  Ictal neocortical slowing in temporal lobe epilepsy.  Neurology. 2004;63(6):1015-1021. doi:10.1212/01.WNL.0000141086.91077.CDPubMedGoogle ScholarCrossref
22.
Englot  DJ, Yang  L, Hamid  H,  et al.  Impaired consciousness in temporal lobe seizures: role of cortical slow activity.  Brain. 2010;133(pt 12):3764-3777. doi:10.1093/brain/awq316PubMedGoogle ScholarCrossref
23.
Spencer  SS, Williamson  PD, Spencer  DD, Mattson  RH.  Human hippocampal seizure spread studied by depth and subdural recording: the hippocampal commissure.  Epilepsia. 1987;28(5):479-489. doi:10.1111/j.1528-1157.1987.tb03676.xPubMedGoogle ScholarCrossref
24.
Kim  JH, Guimaraes  PO, Shen  MY, Masukawa  LM, Spencer  DD.  Hippocampal neuronal density in temporal lobe epilepsy with and without gliomas.  Acta Neuropathol. 1990;80(1):41-45. doi:10.1007/BF00294220PubMedGoogle ScholarCrossref
25.
Spencer  DD, Spencer  SS.  Hippocampal resections and the use of human tissue in defining temporal lobe epilepsy syndromes.  Hippocampus. 1994;4(3):243-249. doi:10.1002/hipo.450040303PubMedGoogle ScholarCrossref
26.
Barba  C, Barbati  G, Minotti  L, Hoffmann  D, Kahane  P.  Ictal clinical and scalp-EEG findings differentiating temporal lobe epilepsies from temporal ‘plus’ epilepsies.  Brain. 2007;130(pt 7):1957-1967. doi:10.1093/brain/awm108PubMedGoogle ScholarCrossref
27.
Spencer  DD, Gerrard  JL, Zaveri  HP.  The roles of surgery and technology in understanding focal epilepsy and its comorbidities.  Lancet Neurol. 2018;17(4):373-382. doi:10.1016/S1474-4422(18)30031-0PubMedGoogle ScholarCrossref
28.
Dickten  H, Porz  S, Elger  CE, Lehnertz  K.  Weighted and directed interactions in evolving large-scale epileptic brain networks.  Sci Rep. 2016;6:34824. doi:10.1038/srep34824PubMedGoogle ScholarCrossref
29.
Epstein  CM, Adhikari  BM, Gross  R, Willie  J, Dhamala  M.  Application of high-frequency Granger causality to analysis of epileptic seizures and surgical decision making.  Epilepsia. 2014;55(12):2038-2047. doi:10.1111/epi.12831PubMedGoogle ScholarCrossref
30.
Friston  K, Moran  R, Seth  AK.  Analysing connectivity with Granger causality and dynamic causal modelling.  Curr Opin Neurobiol. 2013;23(2):172-178. doi:10.1016/j.conb.2012.11.010PubMedGoogle ScholarCrossref
31.
Jacobs  J, Zijlmans  M, Zelmann  R,  et al.  High-frequency electroencephalographic oscillations correlate with outcome of epilepsy surgery.  Ann Neurol. 2010;67(2):209-220. doi:10.1002/ana.21847PubMedGoogle ScholarCrossref
32.
Tamilia  E, Park  EH, Percivati  S,  et al.  Surgical resection of ripple onset predicts outcome in pediatric epilepsy.  Ann Neurol. 2018;84(3):331-346. doi:10.1002/ana.25295PubMedGoogle ScholarCrossref
33.
Martinet  L-E, Ahmed  OJ, Lepage  KQ, Cash  SS, Kramer  MA.  Slow spatial recruitment of neocortex during secondarily generalized seizures and its relation to surgical outcome.  J Neurosci. 2015;35(25):9477-9490. doi:10.1523/JNEUROSCI.0049-15.2015PubMedGoogle ScholarCrossref
34.
Mesgarani  N, Cheung  C, Johnson  K, Chang  EF.  Phonetic feature encoding in human superior temporal gyrus.  Science. 2014;343(6174):1006-1010. doi:10.1126/science.1245994PubMedGoogle ScholarCrossref
×