Prognostic value of insular lobe involvement in temporal lobe epilepsy: A stereoelectroencephalographic study

Authors

  • Thomas Blauwblomme,

    Corresponding author
    1. Department of Pediatric Neurosurgery, APHP, Hospital Necker, Paris, France
    2. University “Paris Descartes”, Sorbonne Paris Cité, Paris, France
    3. INSERM U836, Institute for Neuroscience, Grenoble, France
    • Address correspondence to Thomas Blauwblomme, Department of Pediatric Neurosurgery, Hôpital Necker, 149 rue de Sèvres, 75015 Paris, France. E-mail: thomas.blauwblomme@nck.aphp.fr

    Search for more papers by this author
  • Olivier David,

    1. INSERM U836, Institute for Neuroscience, Grenoble, France
    2. University Joseph Fourrier, Grenoble, France
    Search for more papers by this author
  • Lorella Minotti,

    1. INSERM U836, Institute for Neuroscience, Grenoble, France
    2. University Joseph Fourrier, Grenoble, France
    3. Department of Neurology, University Hospital, Grenoble, France
    Search for more papers by this author
  • Anne-Sophie Job,

    1. INSERM U836, Institute for Neuroscience, Grenoble, France
    2. University Joseph Fourrier, Grenoble, France
    3. Department of Neurology, University Hospital, Grenoble, France
    Search for more papers by this author
  • Serge Chassagnon,

    1. Department of Neurology, University Hospital, Grenoble, France
    Search for more papers by this author
  • Dominique Hoffman,

    1. Department of Neurosurgery, University Hospital, Grenoble, France
    Search for more papers by this author
  • Stéphan Chabardes,

    1. INSERM U836, Institute for Neuroscience, Grenoble, France
    2. University Joseph Fourrier, Grenoble, France
    3. Department of Neurosurgery, University Hospital, Grenoble, France
    Search for more papers by this author
  • Philippe Kahane

    1. INSERM U836, Institute for Neuroscience, Grenoble, France
    2. University Joseph Fourrier, Grenoble, France
    3. Department of Neurology, University Hospital, Grenoble, France
    Search for more papers by this author

Summary

Purpose

Failure of anterior temporal lobectomy for temporal lobe epilepsy has raised the question of insular cortex involvement in these seizures. Because of difficulties in exploring the insula with invasive electroencephalography (EEG) recordings, only few studies have been performed and this question remains unanswered.

Methods

Here, we studied 17 patients who underwent surgery for drug-resistant temporal lobe epilepsy, explored with intracerebral electrodes, with at least one electrode coplanar to the insula. We analyzed seizure propagation patterns from temporal lobe structures to the insula, and their effect on outcome. We used an objective measure of the epileptogenicity of the insula for individual cases and group analysis between patients who were seizure-free after surgery and the others.

Key Findings

All temporal lobe seizures propagated to the insular cortex, with a shorter propagation delay in the case of mesiolateral temporal lobe seizures, thus supporting the existence of a perilimbic network. Epileptogenicity of the insular cortex was not a prognostic factor for outcome after surgery.

Significance

Insular involvement in temporal lobe seizure is not per se a prognostic factor for surgical outcome. Prognosis may be correlated with larger epileptogenic zones that our stereoelectroencephalography spatial sampling could have underestimated.

In a majority of patients with drug-resistant temporal lobe epilepsy (TLE), surgery has proved to be an effective therapy (Engel et al., 2003). However, not all patients with TLE are seizure-free after surgery, as shown by a meta-analysis where the median proportion of long-term (≥5 years) seizure-free patients was around 66% (Engel et al., 2003; Tellez-Zenteno et al., 2005). Yet, only one in three to four patients undergoing temporal lobe (TL) surgery can be considered as seizure-free without taking antiepileptic drugs (Schmidt et al., 2004), thus clearly demonstrating that the epileptogenic zone is not fully removed in a significant number of cases.

The insular cortex has been identified as one of the potential candidates for remaining epileptogenic tissue (Guillaume & Mazars, 1949; Guillaume et al., 1953; Penfield & Faulk, 1955). However, the absence of prognostic significance of insular epileptiform activity after temporal lobe removal, as well as the finding that partial or complete insular ablation did not provide a better outcome but increased the risk of motor deficit (Silfvenius et al., 1964) explains that the insula was no more investigated for decades. More recently, Isnard et al. (2000) have rekindled interest in the insula using stereotactic intracerebral electroencephalographic (SEEG) recordings, showing that this region was frequently invaded by seizures arising from temporal lobe structures. These data were in accordance with neuroimaging studies showing that fluorodeoxyglucose–positron emission tomography (18FDG-PET) interictal hypometabolism often involved the insula in cases of TLE (Arnold et al., 1996; Bouilleret et al., 2002; Chassoux et al., 2004). They were also in agreement with a recent SEEG study quantifying epileptogenicity of brain structures in patients with various forms of TLE, showing that the involvement of the insula was relatively frequent whatever the type of TLE (Bartolomei et al., 2010). However, the type of temporal lobe seizures that gives rise to an insular involvement remains still in debate, and the prognostic significance of such an insular involvement with respect to postoperative outcome has not been studied.

During the past decade, in some patients undergoing SEEG studies at our institution, insular recordings have been performed by means of electrodes inserted along oblique trajectories in order to improve the spatial sampling of the insular region (Afif et al., 2008a). Proceeding this way, preliminary results have suggested that surgical outcome might be better in patients with TLE without insular involvement than in patients with TLE with insular spread (Afif et al., 2008a).

Therefore, we studied here a group of 17 consecutive patients with TLE who underwent insular recordings as part of an SEEG procedure, with the aim of assessing the electrophysiologic modalities of insular spread during temporal lobe seizures, and to search for the possible impact of different spread patterns on surgical outcome.

Patients and Methods

Inclusion criteria

The 17 studied patients belonged to a cohort of 170 patients who underwent SEEG recordings at our institution during a 10-year period (2001–2010), according to the methodology developed in our group (Kahane et al., 2004). They were included in this retrospective study using the following inclusion criteria: (1) normal magnetic resonance imaging (MRI) (Siemens, Malvern, PA, U.S.A.), or MRI evidence of hippocampal sclerosis; (2) SEEG (Micromed, Treviso, Italy) recordings prior to surgery with at least one oblique intracerebral electrode exploring the insular cortex; (3) at least one SEEG-recorded spontaneous seizure; and (4) surgical resection restricted to TL structures.

Patients' characteristics

The 17 patients all had drug-resistant partial epilepsy suspected to involve the temporal lobe, the operability of which could not be decided on the basis of noninvasive procedures only. These latter included in all cases high-resolution MRI, scalp video-EEG monitoring, and neuropsychological tests. Hemispheric dominance for language was evaluated using functional MRI (fMRI) in 15 cases, and 18FDG-PET was performed in nine cases, the results of which will not be discussed here due to the small number.

A summary of the patients' characteristics is provided in Table 1. Briefly, there were 7 male and 10 female patients, whose mean age at surgery was 32.6 years (range 10–54 years). Mean age at seizure onset was 11.3 years (range 0.6–34 years), and mean duration of epilepsy before surgery was 21.3 years (range 6–45 years). MRI demonstrated signs of hippocampal sclerosis in 13 patients (associated, in one case, with a contralateral arachnoidal cyst unrelated to the epilepsy), and was normal in 3. The remaining patient (no. 9) had a right occipital porencephalic cyst secondary to a traumatic brain injury 18 years before, which was not related to current seizures. All the patients were right-handed. Functional MRI showed a left-sided predominance for language in 14 patients, and a bilateral representation in 1. One patient (no. 4) had a responsive cardiac pacemaker that was implanted after video-EEG recordings for preventing falls associated with ictal asystoles.

Table 1. Patients' characteristics
Patient/gender/language lateralizationAge at onset (years)Age at surgery (years)Initial precipitating injuryMRI18-FDG PET hypometabolismSurgeryHistologyEngel's classFollow-up (months)
  1. TLE, temporal lobe epilepsy; F, female; M, male; L, left; R, right; FS, febrile seizure; TBI, traumatic brain injury; HS, hippocampal sclerosis; NA, not available in cases of temporal lobe disconnection; ND, not done; NS, nonspecific; ATL, anterior temporal lobectomy; ATD, anterior temporal disconnection; T1, superior temporal gyrus; T2, midtemporal gyrus; TP, temporal pole; FCD I and FCD III, focal cortical dysplasia type I and type III, according to ILAE classification (Blümcke et al., 2011).

  2. a

    Incomplete disconnection followed by a tailored resection, with no benefits in outcome.

  3. b

    Follow-up until a vagus nerve stimulator was inserted, 16 months after surgery.

  4. c

    Patient died 3 months after surgery of unrelated cause.

1/F/LH1241Meningitis

R HS

L arachnoidal cyst

NDR ATDNAIA36
2/F/LH226FSL HSNDL ATLaNSIII11
3/M/LH651MeningitisL HSL temporalL ATDNAIA19
4/M/LH240FSL HSNDL ATLHSIA43
5/F/LH0.612FSNL temporo-fronto-insularL ATLHSIV16b
6/M/LH614Meningo encephalitisNR TP + T1 + T2R T1 resectionFCD IbIV89
7/F/LH342FSR HSR mesiotemporalR ATLHSIB81
8/F/LH619FSL HSL mesiotemporalL ATLHSIA77
9/M/LH2437TBIR occipital porencephalic cystNDR ATDNAIA58
10/F/LH2554R HSR frontotemporalR ATLHSIB28
11/F/LH2140FSL HSNDL ATLHSIA74
12/M/LH1523EncephalitisL HSL temporalL ATLGliosisIA86
13/M/bilateral310FSL HSR temporalR ATLHSIA35
14/M/LH342L HSR mesiotemporalR ATDNAIA74
15/F/LH3442MeningitisL HSNDL ATLHSIA80
16/F//LH1723NNDL ATLNSID3c
17/F/LH1239FSR HSNDR ATLFCD IIIbIB28

SEEG investigation

In this context of TLE, the SEEG study was judged necessary for lateralizing TL seizure onset, distinguishing between medial and lateral TL seizure onset, and differentiating TL from “temporal plus” or from extratemporal lobe seizure onset (Morris et al., 2008). Particularly, the insular cortex was investigated because electroclinical evidence suggested either a possible insular onset (Isnard et al., 2004; Ryvlin, 2006; Nguyen et al., 2009) or an early perisylvian spread of the seizures during video-EEG monitoring (Bancaud, 1987; Hausser-Hauw & Bancaud, 1987; Isnard et al., 2000; Kahane et al., 2001; Barba et al., 2007).

Intracerebral electrodes were implanted in stereotactic conditions. Preoperative targeting was performed using three-dimensional T1 brain MRI computed with a stereotactic software (VoximR; IVS Solution, Chemnitz, Germany), and a stereotactic and stereoscopic digitalized arteriography to determine avascular trajectories of the electrodes. Insertion of the electrodes (DIXI Médical, Besançon, France; diameter of 0.8 mm; 10–18 contacts, 2 mm length, 1.5 mm apart) was guided by a robotic arm (Neuromate, ISS, Grenoble, France) that was connected to the stereotactic frame and driven by the stereotactic software. A total number of 223 electrodes were implanted in the 17 patients (range 10–15/patient), without complications except in one patient (no. 6) in whom electrodes were removed prematurely because of cutaneous infection. One (in 12 patients) or two (in 5 patients) insular electrodes were inserted through a frontal or parietal twist drill in a triple oblique plane coplanar to the insula (Fig. 1, Table S1), as described previously by our team (Afif et al., 2008a). Other electrodes implanted orthogonal to the midsagittal plane reached the insula via a transopercular route, as described in detail (Frot & Mauguiere, 1999). Overall, the insular cortex was investigated by 184 electrode contacts (range 4–18/patient, see Table S1). In addition, most limbic and paralimbic areas, as well as juxtatemporal regions, were also investigated so that ictal discharges could be assessed into a large temporolimbic network including the insula.

Figure 1.

Postimplantation imaging of patient 13. (AC) Coronal and axial T1-weighted MRI; (B) parasagittal T2-weighted MRI; (D) implantation scheme, drawn after the front and lateral teleradiography in the Talairach and Tournoux spaces. Twelve electrodes were implanted orthogonally to the midsagittal plane, and two oblique electrodes (white arrow, electrode X and Y) are coplanar with the insula. Electrode contacts are identified as hypointensity on the MRI scan and can therefore be located on the patient's anatomy. VAC, vertical to the anterior commissure; VPC, vertical to the posterior commissure.

Anatomic localization of the electrode contacts was identified on the postoperative teleradiography (frontal and lateral view), and then reported on the preoperative MRI, normalized in the Montreal Neurological Institute (MNI) referential using a homemade toolbox of Statistical Parametric Mapping software (www.fil.ion.ucl.ac.uk/spm/software/spm/) working under Matlab7 (The MathWorks, Natick, MA, U.S.A.). Direct anatomic localization on the postimplantation MRI was possible for only four patients (Fig. 1), because postoperative MRI has been performed in our institution for only the last 3 years.

Intracerebral recordings were conducted extraoperatively in chronic conditions (1–3 weeks) with reduced medication using an audio-video-EEG monitoring system (Micromed) that allowed simultaneous recording of up to 128 contacts, with a sampling rate of either 256 or 512 Hz, and an acquisition band-pass filter between 0.1 and 90 Hz or between 0.1 and 200 Hz, respectively, depending on amplifier capacities at the date of recordings. Depth EEG activity was displayed between contiguous contacts at different levels along the axis of each electrode, and analysis of SEEG traces was done visually to delineate the epileptogenic region. Based on ictal SEEG findings, the epileptogenic zone proved to be right temporal in eight cases and left temporal in nine cases.

Surgery

Surgery consisted of temporal lobe resection in 13 cases, and temporal lobe disconnection in 4 cases as described elsewhere (Chabardes et al., 2008). In 16 cases, temporal resection/disconnection included the entire hippocampus, amygdala, parahippocampal gyrus, and the temporal pole. The posterior limits of the neocortical resection/disconnection varied according to SEEG results. In one patient, the resection was restricted to the superior temporal gyrus.

Pathologic examination, available for the 13 resection procedures, confirmed the diagnosis of hippocampus sclerosis (HS) in 8 of 10 patients in whom MRI found signs of HS, including one case in whom HS was associated with focal cortical dysplasia (FCD) type IIIa (Blümcke & Mühlebner, 2011; Blümcke et al., 2011). In three patients without any temporal lobe abnormalities on MRI, histologic examination revealed signs of HS (n = 1), FCD type I (n = 1), FCD type IIIa (n = 1), or nonspecific changes (see Table 1).

Postoperative outcome was assessed according to Engel's classification (Engel et al., 1996). Mean follow-up after surgery was 49.3 months (range 3–89 months). One patient (no. 16) died of cardiac disease unrelated to the epilepsy a few months after surgery, with a postoperative follow-up of only 3 months.

Ten patients were in class IA, three in class IB, one in class ID, one in class III, and two in class IV.

Data analysis

All the SEEG-recorded spontaneous seizures were reviewed, but neither subclinical seizures nor seizures with unusual secondary tonic–clonic generalization were kept for analysis. Seizure semiology was analyzed according to a working definition of ictal and postictal symptoms (see Table 2), and SEEG data were analyzed both visually and using a quantitative approach of epileptogenicity.

Table 2. Electroclinical characteristics of the patients' seizures
Pts/outcSzrsSOZSOZ SEEG patternSzr dur (s)Insular involv (s)Insular involv (norm)Insular SEEG patternIOZ GIInsular GIAurasVegetative signsSimple motor signsComplex motor signsLOCPostictal deficit
  1. Pts, patients; outc, outcome; szrs, seizures; SOZ, seizure onset zone; szr dur, seizure duration; ins inv, insular involvement; consc imp, consciousness impairment; epig, epigastric sensation; emo, emotional modifications (anxiety, fear, happiness); ceph, cephalic sensation; paresth, paresthesia; aud, auditory hallucinations; gust, gustatory hallucinations; pharyng myocl, pharyngeal myoclonia; DP (R or L), right or left dystonic posture; L vers, left version; GA, gestual (complex elaborate movements) automatisms; VA, verbal automatisms; OAA, oro alimentary automatisms.

1/IA1R MeFD221180.08RD19.5−2.5CyanosisL DP, clonic (pharynx), R VOAA+L motor (face)
2/III2L MeFD4680.17RD84.5Epigastric hands paresthesiaeFlushing
3/IA3L MeFD75150.20RD82.3Hands paresthesiaeOAA+/−
4/IA4L MeFD209850.41FD12.5−1.5CephalicBradycardiaOAA+Aphasia
5L MeHP157530.34FD  CephalicPallor, bradycardiaOAA+Aphasia
5/IV6L MeFD9820.02RD3112L clonic (brachiofacial), L V+L motor (arm)
7L MeFD6340.06RD1910.5+
6/IV8R LatFD111250.23RD  Auditory, cephalicMydriasis, sialorrheaClonic (pharynx), head noddingOAA, VA+
9R LatFD103200.19RD  Auditory, cephalicMydriasis, sialorrheaClonic (pharynx), head noddingOAA, VA+
7/IB10R MeFD100530.53RD21.810.5CephalicL DPGA+Aphasia
11R LatFD1430.21FD  AuditorySpeech arrest
12R LatFD700FD  AuditorySpeech arrestAphasia
13R LatFD720.29FD  AuditorySpeech arrest
14R LatFD8200FD  CephalicL DPOAA, GA+Aphasia
8/IA15L MLFD7090.13RD3016GustatoryOAA+/−
16L MLFD6140.07RD  GustatoryOAA+/−
17L MLFD5050.10RD  GustatoryOAA+/−
9/IA18R MLFD5400FD3725EpigastricOAA, GA+
19R MLFD6100FD  EpigastricOAA, GA, VA+
10/IB20R MeFD62470.76RD303.9Epigastric+/−
21R MLFD70500.71RD  Diffuse warmth+/−
11/IA22L MeHP110310.28FD401.5EpigastricSialorrheaR DP, L VOAA, GA+Aphasia
12/IA23L MeHP126160.13RD814.2EpigastricFlushing, sialorrhea, spittingR DP, R paresis (face)OAA+Aphasia, R motor (face)
24L MeHP98110.11RD  EpigastricSialorrhea, spittingR DP, jargonaphasiaOAA+Aphasia
25L MeHP117210.18RD  EpigastricSialorrhea, spittingR DP, R paresis (face)OAA+Aphasia, R motor (arm)
13/IA26R MeHP67280.41RD10.27Tingling of the palateFlushingOAA, GA
27R MeHP80390.49RD  Tingling of the palateFlushing, hyperpneaOAA, GA
28R MeHP82330.40RD  Tingling of the palateFlushing, hyperpneaOAA, GA
14/IA29R MeHP1961260.64RD89.8L tonic (face), R VOAA, GA+
15/IA30L MeHP164990.60FD10.74.2+/−
16/ID31L MeFD140100.07RD363 R DP, jargonaphasiaOAA, GA+Aphasia
32L MeFD13440.03RD  JargonaphasiaGA+Aphasia
17/IB33R MLFD9660.06RD  Warmth (thorax, shoulders)
34R MLFD76100.13RD  Warmth (thorax, shoulders)Breathless
35R MLFD5070.14RD  Warmth (thorax, shoulders) desire to urinateBreathless
36R MLFD7150.07RD8.70.3Warmth (thorax, shoulders)Breathless

The seizure onset zone (SOZ) was visually defined as the first clear SEEG change that occurred prior to the clinical onset of the seizure. SEEG changes at seizure onset were classified either as a fast synchronizing discharge (FD), characterized by a low voltage fast activity over 20 Hz or a fast discharge of spikes (Fig. 2), or as hypersynchronous onset pattern (HP) of low-frequency (<5 Hz) periodic high-amplitude spikes (Wieser, 2004; Fig. 2). The location of the SOZ was classified in one of the following three categories: (1) mesial type (hippocampus +/− amygdala +/− entorhinal cortex); (2) mesiolateral type (mesial structures + temporal pole, or mesial + lateral TL structures); or (3) lateral type (Bartolomei et al., 1999). The insula was considered to be involved by the epileptic discharge when at least one of the insular recording electrodes exhibited either a FD (Fig. 2), or a rhythmic slow (<10 Hz) discharge (RD) of spikes or sharp waves. Particular attention was paid to the delay between the seizure onset and the insular involvement. This delay was normalized according to the total seizure length, so that the comparison between the seizures was not biased by the seizure duration that largely varied between individuals.

Figure 2.

SEEG recordings of seizure onset in two patients. (A) Patient 9. Early involvement of the insula in the seizure. A low voltage fast discharge is observed at ictal onset on the insular contacts. (B) Patient 13. Late spreading of the seizure in the insula. Arrow indicates insular spreading, with a frequency <10 Hz. Name of the electrodes is indicated on the implantation scheme, the most medial contact of each electrode is plot 1.

Epileptogenicity mapping of the explored brain regions was evaluated using a quantitative approach of SEEG signals that we recently developed (David et al., 2011). Briefly, the method quantifies with a t-value the change of SEEG amplitude in high frequencies by comparing interictal and early ictal periods. Here we defined the frequency band of interest as 60–100 Hz (high gamma) and computed epileptogenicity values using time windows of 10 s duration at seizure onset and 30 s duration during interictal baseline. Statistical mapping of epileptogenicity was performed in three dimensions after local interpolation of SEEG gamma amplitude to produce images superimposed on the patient MRI. When several similar seizures were recorded for a patient, we performed a fixed-effect analysis of all seizures to obtain a single value of epileptogenicity. We also produced group maps of epileptogenicity by distinguishing patients with Engel's score equal to IA and the other patients, according to the approach described in David et al. (2011). For detailed group analysis of insula responses between patients, each electrode contact within the insula was associated with a t-value of epileptogenicity by reporting the value of the closest voxel of the t-map of epileptogenicity.

Statistical analysis

They were performed with SPSS20 for Mac (IBM, Armonk, NY, U.S.A.). Significance was set at the 0.05 level.

Modality of insular propagation of the seizures

All the relevant seizures of the 17 patients were included for electrophysiologic analysis in order to gain statistical power. Insular SEEG patterns (FD, RD) were analyzed with respect to SOZ localization (mesial, mesiolateral, lateral) and SEEG pattern at seizure onset (FD, HP) using the chi-square test. Kruskall-Wallis test or Mann-Whitney test were used to analyze the normalized insular propagation delay and the epileptogenicity value in the insula with respect to insular SEEG pattern, SOZ localization, and SEEG pattern at seizure onset.

Outcome

One typical (recognized by the patients or the patient's family as representative of the patient's epilepsy) seizure per patient was kept for analysis to avoid biases, because the number of seizures per patients ranged between 1 and 4. Patient 7 was excluded from the statistical analysis for outcome because she had two different types of seizures (mesial and lateral), and removal of the epileptogenic zone was not achieved on purpose, to preserve language function.

Among the 16 remaining patients we compared the seizure-free group (Engel class IA) versus the others (Engel class Ib–IV). We used Fisher exact test to analyze outcome according to insular SEEG pattern, and Mann-Whitney test to analyze outcome according to normalized insular propagation delay and insular epileptogenicity value.

Results

A total number of 36 seizures were analyzed (range 1–4/patient), the main characteristics of which are reported in Table 2. The SOZ was mesial in 20 cases, mesiolateral (ML) in 10, and lateral in 6, with mean seizure duration of 92.4 s (range 7–221). SEEG pattern at seizure onset consisted of FD in 26 seizures (mesial 10, mesiolateral 10, lateral 6), and of HP in 10 seizures (mesial 10).

Insular involvement during TL seizures

Insular involvement was noticed in all the 36 seizures, with a mean delay of insular spread of 23.6 s (range 0–126), a mean normalized delay of 0.23 (range 0–0.76), and a mean epileptogenicity value of 7.3 (range −2.5 to 25).

Insular SEEG patterns consisted of RD in 26 cases, and of FD in 10. Insular RD tended to be more frequently observed in mesial (16/20) or ML (8/10) TL seizures than in lateral TL seizures (2/6), but this did not reach statistical significance (chi-square test, p = 0.066). The SEEG onset pattern did not influence insular SEEG patterns (Fisher exact test, p = 0.580).

The normalized delay of insular involvement correlated with two factors: the SOZ location and the SEEG pattern at seizure onset (see box plots, Fig. S1). The insula was invaded later in mesial TL seizures than in mesiolateral TL seizures (Mann-Whitney test, p = 0.023), and when the SEEG onset pattern consisted of hypersynchronizing pattern (Mann-Whitney test, p = 0.008). These results were also found with the absolute values of the delay of insular involvement for SOZ location (p = 0.005) and SEEG pattern at seizure onset (p = 0.001).

Epileptogenicity value of the insula was not significantly associated with the location of the ictal onset zone (Mann-Whitney test, p = 0.086), nor with the histologic examination (Mann-Whitney test, p = 0.744). Nevertheless, there was a trend for mesiolateral TL seizures to have a higher insular epileptogenicity than mesial TL seizures (median of the epileptogenicity value: 12 and 3.5, respectively).

There was no correlation between the duration of the epilepsy and the epileptogenicity value in the insula (Spearman correlation test, r = −0.359, p = 0.172).

Relationships between ictal clinical signs and insular involvement

Seizure semiology of each seizure is described in Table 2. Altogether, auras of various types were experienced in most of the seizures (31/36, 86%), vegetative signs were observed in 16 cases (44%), and motor signs were seen in a majority of the cases (27/36, 75%). These latter consisted either in simple motor signs (4/36), complex motor signs (11/36), or both (12/36). Consciousness was impaired at different times relative to SEEG seizure onset in 25 cases (69%). A postictal deficit was noticed in 14 cases (39%), either verbal (10/36), motor (3/36), or both (1/36).

We found no correlation between the epileptogenicity values in the insular cortex and the types of symptoms that were observed during the seizures.

Impact of insular involvement on surgical outcome

Ten of the 16 patients analyzed for surgical outcome were seizure-free (Engel class Ia) after surgery.

Neither insular SEEG pattern of discharge nor normalized delay of insular involvement was associated with the outcome after surgery (Fisher's test, p = 0.585 and Mann-Whitney test, p = 0.536, respectively; Fig. S1).

The absolute values of delay of insular involvement tended to be longer in seizure-free patients than the others (39 vs. 8 sec). Yet, this did not reach statistical significance (Mann-Whitney test, p = 0.055).

Histology was not predictive of outcome (p = 0.108). The ictal onset zone (mesial vs. mesiolateral vs. lateral) was not a prognostic factor (chi-square test, p = 0.274; Fig. 3).

Figure 3.

Epileptogenicity maps. Epileptogenicity maps at seizure onset of the 17 patients reported in the MNI template (p < 0.05, FWE corrected). Patient 7 is represented twice because she had mesial temporal lobe seizures and lateral temporal lobe seizures.

Epileptogenicity values of the insula were not statistically different between group Ia and the others (Mann-Whitney test, p = 0.606; Figs. 4 and S1).

Figure 4.

Group probability maps of epileptogenicity. (A) Group Engel class IA. Mean probability in the insula is 0.4568. (B) Group Engel class Ib–IV. Mean probability in the insula is 0.4055. (C) All patients (except patient 7). Mean probability in the insula is 0.4823.

Group probability maps of having a significant (p < 0.05, family wise error [FWE]) epileptogenicity value at seizure onset were computed for the whole brain (Fig. 4). Clearly temporal and insular structures were found to be associated with high probability. Using an inclusive masking of the insula on epileptogenicity maps, we were able to compute specifically the probability of showing high frequency activities at seizure onset in the insula, which did not show large difference between groups (Engel Ia, p = 0.4568; Engel Ib–IV, p = 0.4055).

Discussion

Significant advances have been made during the last decade for the delineation of the spectrum of insular lobe epilepsy, such as the recognition of different insular seizure types (Isnard et al., 2004; Ryvlin, 2006; Nguyen et al., 2009), the better knowledge of the clinical signs produced by insular stimulation (Ostrowsky et al., 2000, 2002; Isnard et al., 2004; Afif et al., 2008b; Nguyen et al., 2009; Afif et al., 2010), and the demonstration that insular lobe surgery could be safe and effective (Lang et al., 2001; Hentschel & Lang, 2005; Duffau, 2009; Malak et al., 2009). Nevertheless, insular spreading in TL seizures has rarely been evaluated (Isnard et al., 2000; Afif et al., 2008a), and its consequence on surgical outcome remains debated.

The aim of the present study was to address this issue by using an original SEEG methodology that allowed a large sampling of the insular cortex (Afif et al., 2008a), paying particular attention to intracranial electrophysiologic findings, and using a new quantitative evaluation of the epileptogenicity of brain structures.

Our study has limitations: there are only 17 patients, but indications of insular recording are scarce, and only few previous studies provided extensive sampling of the insular cortex. Four patients have a follow-up of <2 years (3, 11, 16, and 19 months, respectively), which may not be enough to conclude on a seizure-free state. There is a selection bias in our study, as insular explorations have been decided because of ictal clinical signs and surface EEG findings suggesting insular involvement in the seizures; therefore, our series may not be representative of all patients with TLE.

Our results confirm first that insular involvement during temporal lobe seizures is common (36/36 seizures), as demonstrated previously by Isnard et al. (2000), who found that 79 of 81 SEEG recorded seizures (in 21 patients) propagated to the insula. In two cases, the seizures were initiated in the insula. More recently Nguyen et al., 2009 found that among 18 patients with frontal, parietal, or temporal lobe seizures studied by depth electrodes, the insula was part of the ictal onset zone in 4 in whom it was subsequently resected because the seizures arose simultaneously from the insula and periinsular cortex. Insular involvement during mesial temporal lobe seizures has also be shown recently by a single photon emission computed tomography (SPECT) study (Sequeira et al., 2013), in which the authors found a decrease in correlation between temporal epileptic cortex and remaining cortex in the interictal state, followed by an increase in the cross-correlated perfusion of the temporolimbic structures (including the insula), corresponding to local network integration.

A second finding relates to the type of TL seizure that could give rise to preferential insular involvement. We showed that the normalized delay of insular involvement was shorter in cases of mesiolateral TLE, and when the SEEG pattern at seizure onset in TL structures consisted of a fast synchronizing discharge. This is in line with previous SEEG data (Chabardes et al., 2005) that showed different spread patterns depending on the initial involvement of the hippocampus or the temporal pole. Indeed, temporal pole seizures propagated to the perisylvian cortex in 87% of the cases, as compared to only 44% of the cases of hippocampal seizures (Chabardes et al., 2005). This could be explained by anatomic and phylogenic background. Indeed, the insular cortex is phylogenetically the pivot of the paralimbic cortex composed of the temporal pole, orbitofrontal cortex, and insula (Mesulam & Mufson, 1982a). These structures share a common cytoarchitectonic organization with rings of agranular, dysgranular, and granular cortex centered by the prepiriform cortex and are strongly interconnected (Mesulam & Mufson, 1982b; Mufson & Mesulam, 1982). On the other hand the mesial structures (amygdala, hippocampus, and entorhinal cortex) send fewer and less direct afferences to the insula (Augustine, 1996; Dupont et al., 2003), thus explaining that the insular propagation of the mesiotemporal seizures was slower in our study.

Owing to the previous results, we looked if the difference of semiology between mesiolateral/temporopolar (Maillard et al., 2004; Chabardes et al., 2005) and purely mesial TL seizures could be explained by insular involvement. But we did not find specific clinical signs, associated with insular spreading of the seizure, neither using visual electrophysiologic analysis nor using the quantification of amplitude changes of high frequency activities (David et al., 2011). This result is in accordance with Isnard et al. (2000).

The last important finding of our study is that insular spread of temporal lobe seizure does not appear as a prognosis factor after temporal lobectomy. This was already noticed by Isnard et al. (2000), who reported that all but 2 of their 21 patients were seizure-free after anterior temporal lobectomy, despite insular spreading of the seizure. Nevertheless, this study had biases, as three patients with an early insular propagation of the seizure (200–400 msec) and a low voltage fast activity (two cases) recorded in the insula were not operated on because of bilateral independent foci in the temporal lobe (two cases) or risk of severe verbal memory impairment (one case). In another study (Afif et al., 2008a), the authors observed a difference in outcome between six patients with TLE with no insular involvement and 11 patients with insular propagation of the seizure (83% vs. 72% of Engel class Ia patients). Yet, this did not reach statistical significance. It is possible that failure of ATL for TLE could be explained by a more widespread network, involving the perisylvian cortex (Ryvlin & Kahane, 2005) but not the insular cortex per se? Indeed, in a recent study using the epileptogenic index (EI; Bartolomei et al., 2008) the authors showed that the prognosis after temporal lobe surgery was worse when the EI was high in extratemporal cortex. Moreover, the number of extratemporal lobe structures (frontoorbital region, operculum, and insula) displaying high EI inversely correlated with the outcome, thus supporting the hypothesis of an epileptic temporal network including both the insula and other perisylvian structures (Bartolomei et al., 2010).

Eventually, our study supports the reliability, and usefulness of objective markers of brain epileptogenicity (David et al., 2011). We could perform group analysis of epileptogenicity mapping in order to highlight the most stable electrophysiologic patterns for patient with the same seizures. As such, probability maps of group epileptogenicity enabled us to perform comparison between class Ia patients and the others, showing once more that the insular cortex solely could not explain failure of surgery.

Conclusion

Insular involvement in TL seizures is the rule and varies according to the location and discharge pattern of the seizure-onset zone. It has no obvious prognostic value for the outcome after temporal lobectomy.

Disclosure

None of the authors have any conflict 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