A longitudinal study of surgical outcome and its determinants following posterior cortex epilepsy surgery


Address correspondence to Lara Jehi, MD, Epilepsy Center, Neurological Institute, Cleveland Clinic, 9500 Euclid Ave, Cleveland, OH 44195, U.S.A. E-mail:jehil@ccf.org


Purpose: To investigate the longitudinal seizure outcome and identify potential prognostic indicators following posterior cortex epilepsy (PCE) surgery.

Methods: We reviewed patients who underwent a parietal, occipital, or parietooccipital resections between 1994 and 2006, using survival analysis and multivariate regression with Cox proportional hazard modeling. A favorable outcome was defined as Engel Class I at last follow-up.

Results: Fifty-seven patients were identified with a mean follow-up of 3.3 years (range 1–12 years). The estimated chance of seizure freedom (SF) was 73.1% at 6 postoperative months, 68.5% at 1 year, 65.8% at between 2 and 5 years, and 54.8% at 6 years and beyond. Most recurrences (75%) occurred within the first 6 postoperative months. Parietal resections had a worse outcome than occipital or parietooccipital resections (52% SF vs. 89% and 93%, respectively, at 5 years). Independent predictors of recurrence included an epilepsy etiology other than tumor or dysplasia [risk ratio (RR) 2.29], limiting resection to a lesionectomy (RR 2.10), having ipsilateral temporal spiking on preoperative scalp electroencephalography (EEG) (RR 2.06), or any ipsilateral spiking on postoperative EEG (RR 2.70) (Log likelihood-ratio test p < 0.0001). Only 40–50% of patients with a poor outcome predictor were SF at 5 postoperative years as opposed to about 80% otherwise. In surgical failures, recurrent seizure frequency was related directly to baseline seizure frequency and to the presence of ipsilateral spiking on postoperative EEG.

Discussion: These data highlight favorable long-term outcomes following PCE surgery. Limited surgical resection and diffuse baseline epileptogenicity may be important predictors of seizure recurrence.

Despite major advances in diagnostic and surgical techniques, parietooccipital resections still represent <10% of all epilepsy surgeries (Barba et al., 2005; Blume et al., 2005; Dalmagro et al., 2005). Reported success rates vary from 25–90% (Blume et al., 1991; Salanova et al., 1992, 1995a,b; Williamson et al., 1992a,b; Cascino et al., 1993; Aykut-Bingol et al., 1998; Bautista et al., 1999; Boesebeck et al., 2002; Barba et al., 2005), leaving this patient category in an undefined “outcome group” between the traditionally considered more “favorable” prospects of anterior temporal lobectomy (Yoon et al., 2003; McIntosh et al., 2004; Jeha et al., 2006) and the frontal lobe surgeries with a relatively poor prognosis (Janszky et al., 2000; Jeha et al., 2007).

Many prior studies were descriptive, reporting on rates of seizure freedom (SF) following parietal (Williamson et al., 1992a; Cascino et al., 1993; Salanova et al., 1995a, 1995b) or occipital lobe resections (Williamson et al., 1992b; Kuzniecky et al., 1997) without identifying specific prognostic outcome indicators. Considering well-described difficulties in drawing anatomic and neurophysiologic distinctions between the occipital and parietal lobes (Sveinbjornsdottir & Duncan, 1993), other studies analyzed a single group of “posterior cortex” surgery limiting the ability to distinguish variations in surgical outcome and its predictors in cases with well-defined pathology in either lobe (Bautista et al., 1999; Boesebeck et al., 2002; Dalmagro et al., 2005). All previous reports were cross-sectional. Therefore, none evaluated longitudinal aspects of recurrence, or investigated potential prognostic predictors using modern statistical techniques designed for time-dependent outcomes.

Our goal is to describe the long-term seizure outcome and its predictors following parietal, occipital, and multilobar resections in the posterior cortex. We use the statistical methods of survival analysis and proportional hazard modeling to evaluate rate, stability, and predictors of SF while accounting for variation in the duration of follow-up among patients. In addition, we will briefly address the issue of postoperative complications.


Patient selection

We reviewed patients of all age groups with medically refractory posterior cortex epilepsy (PCE) who underwent resective surgery at Cleveland Clinic between January 1994 and January 2006, excluding those with prior brain surgery. Follow-up of at least 1 year was required. Fifty-seven patients fulfilled all inclusion and exclusion criteria.

Preoperative patient evaluation protocol

Noninvasive testing

All patients had a detailed neurologic examination and prolonged video-EEG (electroencephalography) monitoring with scalp electrodes placed according to the International 10–20 system, with sphenoidal electrodes used in most adults. Interictal discharges and ictal patterns were analyzed by lobar distribution (parietal, occipital, frontal, or temporal), relationship to the side of surgery (ipsilateral or contralateral), and their extent (regional, lateralized, or generalized). A lateralized ictal pattern was defined if both ipsilateral electrode chains were involved and the amplitude of the contralateral hemisphere pattern was less than 50%. If a clear side of onset was not present, or if the initial EEG change was diffuse, the seizure was defined as non-localizable or generalized.

Clinical seizure semiology was recorded (Luders et al., 1998), as well as the occurrence of lateralizing seizures or auras. Acute postoperative seizures (APOS) as defined by seizures occurring within the first postoperative week were also recorded. Formal visual field testing was performed in the majority of patients before and after surgery.

All patients had high-resolution magnetic resonance imaging (MRI) performed with a Siemens 1.5 Tesla SP system (Siemens, Erlangen, Germany) using a standardized epilepsy protocol that included high-resolution T1-weighted volume acquisition, T2-weighted, and fluid attenuated inversion recovery (FLAIR) sequences. Most also had a fluorodeoxyglucose (FDG)–positron emission tomography (PET) scan. Imaging results were classified as either being normal, or showing unilobar or multilobar abnormalities. Results of the noninvasive evaluation were discussed in a multidisciplinary patient management conference where the decision was made to either proceed with a resection, or to perform an invasive evaluation for further investigation.

Invasive EEG

Subdural electrodes (SDEs) were implanted in 24 patients (42% of cases) for any of the following reasons: (1) to confirm the ictal onset zone if scalp EEG information was either nonlocalizing or inconsistent with other noninvasive investigations such as MRI or seizure semiology; (2) to perform functional mapping preoperatively if the ictal onset zone was thought to be close to eloquent cortex; and (3) whenever a multifocal epileptogenic process was suspected based on multifocal noninvasive EEG findings or multilobar involvement on imaging. Details of the SDE insertion, MRI coregistration, and invasive EEG analysis are described in detail elsewhere (Widdess-Walsh et al., 2007).

Nineteen cases had intraoperative electrocorticography (ECoG) to either confirm removal of epileptogenic cortex (in five), perform awake functional cortical electrical stimulation mapping (in four), or do both (in 10). All awake craniotomies occurred with surgeries involving the parietal lobe, with somatosensory evoked potential (SSEP)–based central sulcus localization and language mapping being the main goals.

Surgery and pathologic subgroups

Surgeries were classified by location (parietal, occipital, or parietooccipital) and by type (lesionectomy, lobectomy, or multilobar resection). Etiologic subgroups, as based on MRI and surgical pathology, were: (1) Malformations of cortical development (MCDs) with abnormal MRI; (2) MCDs with normal high-resolution MRI; (3) vascular malformations; (4) cryptogenic with normal MRI and histology; (5) encephalomalacia following stroke or trauma; and (6) tumors. We included only low grade tumors with refractory epilepsy. Surgery was, therefore, performed to primarily treat the seizures rather than the tumor, with lesionectomy being performed in 70% of tumors, and the rest undergoing lobectomy or multilobar resections when preoperative testing suggested more diffuse epileptogenicity.

Postoperative patient evaluation protocol

All patients underwent a postoperative high-resolution MRI. Completeness of resection was determined based on removal of the MRI lesion. In nonlesional cases, resection was considered complete if there was complete removal of the brain tissue covered by the electrodes of ictal onset and most frequent interictal abnormalities defined by SDE/intraoperative ECoG evaluations.

All patients had a routine outpatient EEG and a visual field examination at six postoperative months.

Outcome definition

Patients were classified as seizure-free if they maintained an Engel score of I at their last follow-up. Postoperative clinical information was obtained from patient visit notes and follow-up phone calls. The typical follow-up schedule consisted of a clinic visit at three postoperative months, 6 months, 1 year, and then yearly. The primary outcome was time to first seizure recurrence. In patients who had APOS within the first postoperative week, the time to the first recurrence following the APOS was considered. When medical records indicated that a recurrence occurred between two follow-up visits or phone contacts, the date of recurrence was taken as the midpoint of this period. With all patients having the follow-up schedule detailed previously, the period between documented contacts (visits or phone calls) never exceeded 1 year.

Statistical methods

Prior to modeling, the data were summarized with descriptive statistics for each variable. For exploratory purposes, an initial univariate analysis was performed using Wilcoxon rank-sum, chi-square, and Fisher’s exact tests to compare seizure-free patients to those with seizure recurrence. This allowed identification of potential prognostic indicators, which were subsequently tested in a multivariate Cox proportional hazards regression model. Results were considered statistically significant at the 5% level. This method allows testing the correlation of specific variables with outcome while taking into account any interaction among those variables, and their variation with time.

Kaplan-Meier survival analysis was used to calculate the probability of SF. Statistical significance was tested using the log-rank test and comparison of 95% confidence intervals (CIs).


Patient characteristics

Among the 57 patients analyzed, resections were occipital in nine, parietal in 32, and parietooccipital in 16. The majority were lesionectomies (27 cases; 48%), followed by lobectomies in 19 cases (33%) and multilobar resections in 11 (19%). The mean follow-up duration was 3.3 years ± (SD) 2.7 years (median 3.1; range 1–12 years). At last follow-up, 38 patients (67%) were seizure-free.

Tables 1 and 2 summarize the patients’ baseline clinical characteristics and seizure semiology, whereas Table 3 describes their imaging findings. Table 4 summarizes the case distribution by etiology.

Table 1.   Clinical characteristics of the overall cohort, with comparison between the seizure-free and non–seizure-free groupsa
 Overall group (N = 57)Seizure-free (N = 38)Recurred (N = 19)p-value
  1. ap-value is for Wilcoxon rank-sum, chi-square, or Fisher’s exact tests as appropriate.

  2. GTC, generalized tonic–clonic.

Clinical characteristics
 Female N (%)23 (40%)16 (42%)7 (37%)0.70
 Left-sided surgery35 (62%)22 (58%)13 (68%)0.44
 Mean age at onset, years (range)9.2 (0–40)7.7 (0–38)12.4 (0–40)0.12
 Mean age at surgery, years (range)20.4 (0.3–55.4)18.1 (0.3–47.1)24.9 (1.2–55.4)0.09
 Mean preoperative seizure Frequency/month (range)48 (0.3–600)52 (0.5–600)38.2 (0.33–180)0.57
 Mean epilepsy duration, years (range)10.8 (0.15–48.3)10.4 (0.15–48.3)11.6 (0.5–43.4)0.69
 Preoperative GTC present N (%)39 (69%)26 (72%)13 (68%)0.77
 Preoperative auras present N (%)34 (65%)24 (67%)10 (63%)0.77
 Positive family history N (%)7 (13%)2 (12%)5 (14%)0.83
Table 2.   Semiologic characteristics of the overall cohort, with comparison between the seizure-free and non–seizure-free groupsa
 Overall group (N = 57)Recurred (n = 19)
  1. aN and n equal number of patients in each group; p-value is for chi-square test.

Aura characteristics
Type (p = 0.83)
 Psychic128 (67%)
 Auditory21 (50%)
 Visual123 (25%)
 Sensory61 (17%)
 Nonspecific21 (50%)
Lateralizing aura (p = 0.65)
 Yes82 (25%)
 No268 (33%)
Seizure characteristics
Lateralizing seizure (p = 0.27)
 Yes4413 (30%)
 No136 (46%)
Immediate postoperative seizures (p = 0.79)
 Yes176 (33%)
 No3812 (32%)
Table 3.   Radiologic characteristics of the overall cohort, with comparison between the seizure-free and non–seizure-free groupsa
 Overall group (N = 57)Recurred (n = 19)
  1. aN and n equal number of patients in each group; p-value is for chi-square test.

  2. MRI, magnetic resonance imaging; PET, positron emission tomography.

MRI characteristics
Abnormal (p = 0.59)
 Yes5517 (31%)
 No21 (50%)
Multilobar abnormalities (p = 0.59)
 Yes93 (33%)
 No4614 (30%)
Hippocampal atrophy (p = 0.75)
 Yes41 (25%)
 No5217 (33%)
PET characteristics
Focal abnormality (p = 0.70)
 Yes3911 (28%)
 No41 (25%)
Table 4.   Distribution by disease etiology of the overall cohort, with comparison between the seizure-free and non–seizure-free groupsa
 Overall group (N = 57)Recurred (n = 19)
  1. aN and n equal number of patients in each group.

  2. DNET, dysembryoplastic neuroepithelial tumor; MCD, malformation of cortical development.

MCD (abl MRI)17 (30%)3 (18%)
MCD (nl MRI)2 (4%)1 (50%)
Tumor22 (39%)7 (32%)
 Low-grade glioma72
 Low-grade astrocytoma21
Vascular malformation3 (5%)2 (67%)
Cryptogenic1 (2%)1 (100%)
Encephalomalacia11 (20%)5 (45%)

Forty percent of the patients were female, and 62% had left-sided surgery. Thirty-four patients had auras, which were lateralizing in 24%. Seizure semiology was lateralizing in 77%. Most cases were lesional: Only one patient had a normal MRI, and 39 of 43 (91%) of PET scans performed were abnormal. Three patients had homonymous hemianopsia, and two had contralateral weakness preoperatively.

On scalp EEG, interictal spiking was more concentrated in the occipital (in 40% of cases) and parietal (60%) distributions, but was also seen in the ipsilateral temporal (45%) and frontal (33%) head regions. Contralateral interictal spiking was seen most often in the temporal distribution (24%), followed by the contralateral parietal (15%), contralateral frontal (13%), and least often in the contralateral occipital (7%) head regions. Ictal patterns lateralized to the side of surgery in most cases (91%). Seven patients had some EEG seizures that were contralateral to the side of surgery, and 10 had nonlocalizable seizures.

During SDE evaluations, random spiking (>7 Hz, >200 μV, irregular firing) constituted the most frequently observed interictal activity (18 of 24 cases), whereas slow repetitive spikes (>0.5 s, >200 μV, regular firing) were seen in seven cases, and two had paroxysmal fast discharges (>10 Hz, >0.5 s, regular firing). This interictal spiking was mostly focal (19 of 24 cases), with a mean of three foci per patient in the remaining cases, with more widespread but still regional interictal activity. The most commonly observed ictal onset rhythm was paroxysmal fast activity (in 57% of patients), followed by repetitive spiking in 33%, and diffuse suppression in 10%. A single ictal onset zone (Widdess-Walsh et al., 2007) was found in 41% of monitored patients, with two zones being found, on average, in the rest. Foci of interictal spiking overlapped with ictal onset zones in all but one case.

Resection was felt to be complete in 69%. Acute postoperative seizures occurred in 17 patients (30%), six of whom had ongoing seizures by the last follow-up (33%) as compared to 12 of 38 patients without APOS (p = 0.79). Findings of the routine EEGs performed on patients still seizure-free at 6 months are summarized in Table 5.

Table 5.   Routine EEG findings at 6 months after surgery in patients who were still seizure-free up to that point.a
 Overall group (N = 30)Recurred (n = 4)
  1. aN and n equal number of patients in each group; p-value is for chi-square test.

Postoperative EEG characteristics
Ipsilateral spiking (p = 0.03)
 Yes83 (38%)
 No221 (5%)
Contralateral spiking (p = 0.04)
 Yes11 (100%)
 No293 (11%)


New neurologic deficits developed in 16 patients (28%) after surgery. Those consisted mainly of visual field deficits with contralateral quadrantanopsia in three cases and hemianopsia in five. Another eight patients had variable combinations of the following:

New hemiparesis (in one), worsening of baseline hemiparesis (two), new leg monoparesis (one), and hemisensory loss (two). All cases of hemiparesis were seen in parietal resections. Sensorimotor complications resolved within 2–4 months in four cases, with the remaining patients exhibiting permanent deficits.

Aphasia occurred in three cases, and alexia without agraphia in one. Details of noninvasive language testing are shown in Table 6. Available preoperative functional maps (Fig. 1) and postoperative MRIs (Fig. 2) for all those patients who developed language deficits are shown.

Table 6.   Details of the four patients who developed postoperative language deficits
  Age at surgery
WADASubdural electrodes placedExtra-operative language mapping performedIntra-operative cortical language mappingResectionLanguage deficitOther complicationsDuration of language deficit
  1. aElectrodes were not placed for fear of mass effect in this patient with an astrocytoma.

  2. bRemoved the following day secondary to epidural hematoma compressing left frontal, parietal, and occipital lobes.

  3. cOnly expressive language area identified.

  4. NA, not available.

Patient 19Left language dominantNoaNoNoOccipitalAlexia without agraphiaNone1 week
Patient 245NAYesYesNoInferior parietal lobuleAphasiaHand and face hemiparesisImproved to mild reduction in fluency by 6 months
Patient 314Left language dominantYesbNoNoInferior parietal lobuleAphasiaFace, hand, and leg spastic hemiplegiaImproved to mild reduction in fluency by 6 months
Patient 411NAYesYescNoFronto-parietal operculumAphasiaFace, hand, and leg spastic hemiparesisAlmost resolved by 1 week
Figure 1.

 This shows the patient’s own preoperative brain magnetic resonance imaging (MRI) with reconstructions illustrating superimposed functional maps on Patients 2 (PT 2) and 4 (PT 4) of Table 6. Each green dot represents an intracranial electroencephalography (EEG) electrode. The dark red line represents the central sulcus, as identified by somatosensory evoked potentials. The colored circles represent identified eloquent cortex: red, upper extremity motor; green, upper extremity sensory; purple, face motor; yellow, face sensory. Pink circles represent ictal onset zones.

Figure 2.

 Postoperative magnetic resonance imaging (MRI) showing a sagittal view of the left-sided resections in our four patients with postoperative language deficits (figures are labeled with patient numbers as used in Table 6. For example, PT 1 is Patient 1 from Table 6).

Overall recurrence

The Kaplan-Meier survival curve illustrating seizure recurrence in the overall group is shown in Fig. 3. The estimated chance of SF was 73.1% (95% CI 67–79) at 6 postoperative months, 68.5% (95% CI 62–75) at 1 year, 65.8% (95% CI 59–73) between 2 and 5 years, and 54.8% (95% CI 43–66) at 6 years and beyond. The median timing of recurrence was 2.0 months, with 75% of the seizure recurrences occurring by 6.4 months. Late recurrences were rare, with the latest being at 74 months.

Figure 3.

 Kaplan-Meier plot illustrating chances of postoperative seizure freedom (SF) after posterior cortex surgery in the overall cohort.

Table 7 shows the outcomes using Engel’s classification at last follow-up. Of the 19 patients with recurrent seizures, 11 underwent a repeat video-EEG evaluation (results summarized in Fig. 4).

Table 7.   Detailed seizure outcome using Engel classification for the overall cohorta
  1. aN equals number of patients in each group.

  2. bTwo with sleep deprivation and three unprovoked.

  3. cTwo of those only had one seizure.

  4. AED, antiepileptic drug.

Class I: free of disabling seizures
 A: Completely seizure-free since surgery29
 B: Nondisabling simple partial seizures only  since surgery1
 C: Some disabling seizures after surgery,  but free of disabling seizures for at least 2 years5b
 D: Generalized convulsions with AED  discontinuation only3c
Class II: Rare disabling seizures (“almost seizure-free”)5
Class III: Worthwhile improvement7
Class IV: No worthwhile improvement7
Figure 4.

 Illustration of the results of video-EEG (electroencephalography) reevaluations and outcomes of the 19 patients who failed surgery.

Overall, there was no improvement in seizure frequency in patients who did not achieve SF: The mean monthly seizure frequency in surgical failures actually tended to increase, going from 38.2 at baseline to 73.8 postoperatively, although this change was not statistically significant (p = 0.33). Upon closer analysis, this apparent worsening was mostly seen after unsuccessful parietal lobe surgery (postoperative seizure frequency of 91.1/month compared to 45.6 at baseline; p = 0.16), as seizures remained essentially unchanged in failed occipital (0.33/month at baseline vs. 2.0 postoperatively; p = 0.99) and parietooccipital resections (14/month vs. 16; p = 0.56). The frequency of recurrent seizures was related directly to baseline seizure frequency (R2 = 0.44, p = 0.0027) and to ipsilateral spiking on postoperative EEG (mean seizure frequency of 108/month if spikes were seen vs. 3/month otherwise, p = 0.05, Student’s t test).

Univariate analysis

Potential outcome predictors identified on univariate analysis are identified in Table 8. The independence and significance of those variables was assessed in the multivariate proportional hazard model.

Table 8.   Variables correlating with complete postoperative seizure freedom on univariate analysis
 NRecurred N (%) p-value
  1. aIncluding 11 multilobar resections and five resections of lesions at the parietooccipital junction.

 Parietal3215 (47%)0.04
 Occipital91 (11%)
 Parietooccipitala163 (19%)
Extent of resection
 Lesionectomy2712 (44%)0.005
 Lobectomy197 (37%)
 Multilobar resection110 (0%)
Age at surgery (years)
 ≤18317 (23%)0.05
 >182612 (46%)
Completeness of resection
 Complete3910 (26%)0.03
 Incomplete169 (56%)
 Tumor/dysplasia3910 (26%)0.12
 Other178 (47%)
Ipsilateral temporal spiking
 Present2511 (44%)0.10
 Absent327 (23%)
Interictal spiking
 Focal194 (21%)0.12
 Nonfocal32 (67%)
Ictal rhythm
 Suppression/fast152 (8%)0.04
 Repetitive spikes74 (57%)
Ipsilateral spiking
 Present83 (38%)0.03
 Absent221 (5%)
Contralateral spiking
 Present11 (100%)0.04
 Absent293 (11%)

Multivariate analysis

After multivariate proportional hazard modeling, we found that patients at highest risk of seizure recurrence were those with an epilepsy etiology other than tumor or MCD [risk ratio (RR) 2.29 95% CI 1.27–4.25], who underwent a lesionectomy as opposed to lobectomy or multilobar resection (RR 2.10 95% CI 1.20–3.81), had ipsilateral temporal spiking on preoperative scalp EEG (RR 2.06 95% CI 1.20–3.69), or ipsilateral spiking on postoperative EEG (RR 2.70 95% CI 1.55–4.87) (Log likelihood-ratio test p < 0.0001) (Table 9).

Table 9.   Variables correlating with postoperative seizure freedom after applying multivariate proportional hazard modeling
VariableRisk ratio95% CIp-value
  1. aAs opposed to other surgery type (lobectomy or multilobar resections).

Ipsilateral preoperative temporal spiking2.061.20–3.690.0087
Ipsilateral postoperative spiking2.701.55–4.870.0004
Etiology other than MCD or tumor2.291.27–4.250.0061

Seizure-free rates for the identified independent outcome predictors were calculated at 6 months, and 1, 2, and 5 postoperative years [Table 10], and survival analysis curves were constructed [Fig. 5]. The most important predictor of long-term outcome was spiking on the 6-months postoperative EEG (RR 2.70, 95% CI 1.55–4.87, p = 0.0004): Around 80% of patients without spikes at 6 months were still seizure-free 5 years following surgery, as opposed to only about one-third if spikes were seen. The second most significant outcome predictor was the type of surgery, even after controlling for etiology. In “favorable etiologies” such as tumor/MRI-visible dysplasia, all patients with multilobar resections were seizure-free at last follow-up (100%), as opposed to 63% of those with a lobectomy and 67% of those with a lesionectomy (p = 0.04). With the remaining “unfavorable etiologies,” again 100% of multilobar resections became seizure-free as opposed to 64% of those with a lobectomy and 20% of those with a lesionectomy (p = 0.07).

Table 10.   Estimated chances of complete seizure freedom at 1, 2, and 5 postoperative years as obtained from survival analysis by each of the identified prognostic indicators
Seizure free rates6 months 1 year 2 years 5 years
MCD/tumor (p = 0.0061)
  Yes84 ± 6%78 ± 7%74 ± 8%62 ± 13%
  No52 ± 12%52 ± 12%52 ± 12%52 ± 12%
Temporal spiking preoperatively (p = 0.0087)
  Absent83 ± 7%78 ± 8%78 ± 8%59 ± 18%
  Present63 ± 10%58 ± 10%52 ± 11%52 ± 11%
Type of surgery (p = 0.0091)
  Lesionectomy67 ± 9%58 ± 10%58 ± 10%46 ± 13%
  Lobectomy/multi-  lobar resection79 ± 8%79 ± 8%72 ± 9%72 ± 9%
Spiking on 6-month-postop EEG (p = 0.0004)
  YesN/A44 ± 12%37 ± 12%37 ± 12%
  NoN/A84 ± 7%79 ± 8%79 ± 8%
Figure 5.

 Survival analysis showing postoperative seizure outcome depending on the various identified prognostic indicators.

For all the prognostic variables, the difference in outcome between the favorable and unfavorable groups became apparent very early postoperatively: By as soon as 1 year after surgery, about 80% of patients within favorable prognostic indicators were seizure-free as opposed to around 50% otherwise. Over the long-term, this outcome gap narrowed as the chances of SF decreased in all patient groups, but the difference remained significant (Fig. 5).

Subgroup analysis to study the identified prognostic variables by individual locations within the posterior cortex (parietal versus occipital versus parietooccipital) was not statistically possible in view of small patient numbers.

There was a high correlation between the number of favorable prognostic variables present and the chance of SF. Ten patients had all four favorable indicators: All were seizure-free at last follow-up. Of 11 patients who had ≤1 favorable indicator, only 18% were seizure-free. Of the remaining 35 patients with 2–3 favorable indicators, 74% were seizure-free (R2 = 0.32, p = 0.0002).

AED discontinuation

The general course in our center is to simplify the antiepileptic drug (AED) regimen in patients who have been seizure-free for 1–2 years, and have neither residual epileptiform findings on EEG nor potentially epileptic abnormalities on postoperative MRI. This “practice,” however, is very much patient- and physician-dependent. Within the limitations of such an uncontrolled setting, this review identified 13 patients for whom AEDs were discontinued. Seven remained seizure-free off medication (after a mean follow-up of 4.2 years). The remaining six had a breakthrough seizure, and AED treatment was resumed: Three regained SF, and three (3 of 13, 23%) had persistent seizures. Forty-seven patients were still on AEDs at last follow-up. The AED status was unknown in three patients who were seizure-free at last follow-up. There was no correlation between the occurrence of breakthrough seizures in the setting of AED withdrawal and subsequent seizure severity or seizure outcome at last follow-up.


We report here the first longitudinal outcome study describing seizure recurrence and its predictors following PCE surgery.

Overall seizure recurrence

Previously reported success rates following PCE surgery varied from 25–90% (Blume et al., 1991; Salanova et al., 1992; Williamson et al., 1992a, 1992b; Cascino et al., 1993; Salanova et al., 1995a, 1995b; Aykut-Bingol et al., 1998; Bautista et al., 1999; Boesebeck et al., 2002; Barba et al., 2005). Review of the curve of seizure recurrence in our cohort (Fig. 3) shows that the chance of SF is highly dependent on time of last follow-up: Although close to 75% of patients are seizure-free at 6 months following surgery, this number falls abruptly to around 70% by 1 year and then to slightly greater than 50% by 8 years. This longitudinal variation may at least partly explain why previous cross-sectional studies performed at various points of follow-up after PCE surgery may have provided different success rates.

In our PCE surgical cohort, surgical failures were identified as early and as quickly as their frontal lobectomy counterparts, with 75% of all recurrences in our cohort and 80% of those following frontal lobe epilepsy surgery (Jeha et al., 2007) occurring within the first 6 postoperative months. However, our cohort’s long-term SF rates mirrored the more optimistic outlook of temporal lobectomy cases, with close to 50% of patients seizure-free 5–10 years after surgery (McIntosh et al., 2004; Spencer et al., 2005; Jeha et al., 2006). This suggests that it may be over simplistic to lump all neocortical “extratemporal” epilepsies within the same clinical and outcome groups (Sinclair et al., 2004; Yun et al., 2006), and that there might be multiple mechanisms to explain seizure recurrence beyond anatomic factors.

Furthermore, we identified predictors of the severity of seizure recurrence as reflected by postoperative seizure frequency. We and others have previously shown that frequent preoperative seizures (>20/month) predict more frequent postoperative seizures following anterior temporal lobectomy (Khoury et al., 2005; Jeha et al., 2006). Our current study shows that the same applies following PCE surgery. We identified spiking on postoperative EEG as an additional predictor of the severity of recurrence.

Within our PCE cohort, patients with isolated parietal lobe resections had the worst outcome: At last follow-up, about half the patients who underwent parietal lobe surgeries had seizure recurrence, as opposed to only10–20% of those patients who underwent occipital and parietooccipital resections (Table 5, Fig. 6). Furthermore, parietal failures had more frequent postoperative seizures, whereas seizure frequencies following unsuccessful occipital and parietooccipital resections were unchanged (as compared to baseline). However, with multivariable regression analysis, those differences were not statistically significant. This may be due to our small patient numbers, or alternatively to a larger proportion of lesionectomies in the parietal lobe (56% of resections) as opposed to elsewhere in the posterior cortex in our cohort (44% with p = 0.25). This suggests that lower success rates with parietal surgeries may simply reflect more conservative resections of the epileptogenic zone there, probably for greater fear of injury to eloquent (postcentral, or dominant angular/supramarginal gyri) cortex. This would agree with recent work showing that resections of the somatosensory cortex achieve better outcomes when surgery extends to involve the precentral and inferior rolandic cortex (Pondal-Sordo et al., 2006).

Figure 6.

 Kaplan-Meier plot illustrating chances of postoperative seizure freedom in patients with parietal lobe surgery, as opposed to occipital and multilobar resections in the posterior cortex.

Predictors of Seizure Recurrence

We identified four independent predictors of seizure outcome following PCE surgery (Table 9).

Disease etiology

The most significant preoperative outcome predictor identified was disease etiology: Only about one-fourth of patients with either a tumor or lesional MCD had postoperative seizure recurrence, as opposed to more than one-half of the patients who had other pathologies (RR 2.29; Tables 4 and 5). This concurs with previous studies showing a tendency for favorable outcomes in lesional MCDs (Kuzniecky et al., 1997; Dalmagro et al., 2005; Urbach et al., 2007), and more so in tumors (Aykut-Bingol et al., 1998; Boesebeck et al., 2002). This also suggests that similar to low-grade neoplasms, high-grade MRI-visible MCDs (types IIA and IIB) are amenable to successful surgical treatment (Kuzniecky et al., 1997; Hamiwka et al., 2005; Yun et al., 2006; Caicoya et al., 2007; Urbach et al., 2007).

Extent of resection

Similar to a previous PCE surgery series (Barba et al., 2005) we found completeness and extent of resection to be an important outcome predictor: Patients who underwent a lesionectomy were about twice as likely to exhibit seizure recurrence as compared to those who underwent either a lobectomy or multilobar resection (Table 8). In our study, we used the traditional way of completeness of resection assessment through analysis of postoperative MRI. Our observation though that 27% of patients with complete resection of their MRI lesion had spiking on their 6 month routine EEG versus 56% of those with incomplete resections (p = 0.05) suggests that spiking on a postoperative EEG may be another useful indicator of the completeness of the resection. This finding is not surprising but was not previously confirmed (Boesebeck et al., 2002).

Although we would expect it to be easier to perform “complete resections” in the nondominant hemisphere, there was no correlation between side of surgery and completeness of resection (p = 1.0) or with extent of surgery (p = 0.68) in our cohort, suggesting that extent of resection is not solely governed by concern of injury to eloquent cortex. In our cohort, better outcomes were seen with larger resections, regardless of etiology. This would, therefore, support performing more aggressive and larger resections to treat intractable PCE whenever possible. However, our observation that a lesionectomy achieved SF in 67% of cases in tumor or MCD but in only 20% of other etiologies suggests that attempting a “smaller surgery” to avoid injuring eloquent cortex may be appropriate in selected cases of tumor/MCD, while preferably attempting more aggressive and larger resections in patients with nontumor non-MRI lesional MCDs.

Preoperative temporal spiking

In our cohort, patients with preoperative ipsilateral temporal spiking were twice as likely to fail surgery compared to patients with no extraposterior (parietooccipital) cortex spiking (RR = 2.06, 95% CI 1.20–3.69, p = 0.0087). Of the 11 patients with baseline temporal spiking who failed surgery, nearly one-half had neuroimaging abnormalities in the ipsilateral temporal lobe: Four had parietal MRI lesions extending to the posterior temporal region (schizencephaly in two, stroke in one, and perisylvian tumor in one) and two had temporal hypometabolism on PET, despite the lack of a clear temporal lobe MRI lesion. When reevaluated with video-EEG post seizure recurrence, 5 of those 11 patients had seizures arising from the temporoparietal junction. Three refused further surgery, whereas two underwent a temporal lobectomy. One became seizure-free and seizures persisted in the second. Surgical failure in this group may then have been caused by an incomplete resection of a rather large epileptogenic region extending anteriorly to the insula or the anterior temporal lobe. This would support careful evaluation of the need to include the temporal lobe in a PCE surgery when there is evidence of imaging or electrophysiologic abnormalities pointing and/or extending to the temporal lobe.

The negative prognostic value of temporal spiking in our study concurs with previous work showing that interictal spiking extending beyond the area of resection correlates with poor surgical outcome in patients with extrahippocampal epilepsy (Blume et al., 1991; Bautista et al., 1999). This was not reproduced in studies where interictal EEG was analyzed only by its distribution (focal, regional, diffuse, and so on) without accounting for location (temporal, parietal, among others) (Dalmagro et al., 2005) or where only a small number of patients actually had temporal interictal activity (Boesebeck et al., 2002). Temporal spread of occipital seizures has been documented repeatedly, both electrically via intracerebral depth recordings (Takeda et al., 1970; Salanova et al., 1992), and semiologically with anterior propagation of occipital seizures producing automatisms (Collins & Caston, 1979). A negative prognostic implication of this spread, as suggested by previous studies (Blume et al., 1991; Bautista et al., 1999), and the observation of more favorable outcomes following combined occipital and temporal resections as opposed to more focal occipital resections in patients with “occipitotemporal” epilepsy (Aykut-Bingol & Spencer, 1999) may support the existence of a so-called occipitotemporal secondary epileptogenesis (Palmini et al., 1999).

Spikes on postoperative scalp EEG

In our series, only 1 of 22 patients with no spikes on their postoperative EEG subsequently developed recurrent seizures, as opposed to 38% of those patients who did show postoperative epileptic EEG abnormalities (Table 5). In patients who were seizure-free 6 months following temporal (Jeha et al., 2006) or frontal (Jeha et al., 2007) resections, spikes on routine EEG also predicted later seizure recurrence. Our current results illustrate the role and importance of postoperative EEG findings in patients who undergo parietal and/or occipital resections, as they may reflect an incomplete resection of the epileptogenic zone.

Other outcome predictors

Neither performing intraoperative functional mapping nor using SDE correlated with completeness of resection (p = 0.49 and p = 0.39, respectively), presumably because eloquent areas were identified and thus avoided. However, findings on intracranial EEG recording may be relevant in predicting outcome. Similar to previous studies (Bautista et al., 1999; Barba et al., 2005), focal interictal spiking correlated with better outcomes. Although the focalization of ictal onset and the ictal onset frequency in invasive studies were not related to surgical outcome in previous cohorts of neocortical epilepsy (Lee et al., 2005; Widdess-Walsh et al., 2007), we found that fast activity at seizure onset correlated with better outcome (8% recurrence risk vs. 57%) but only on univariate analysis. Since this ictal rhythm was seen in 75% of cases with tumor or dysplasia as opposed to 60% of patients with other disease etiologies in our cohort (p = 0.10), it is possible that it simply represents a surrogate marker for disease etiology, rather than an independent outcome predictor.

Similar to previous studies, we found no correlation of side of surgery, gender, presence or absence of auras, clinical semiology, or surgical outcome (Blume et al., 1991; Salanova et al., 1992; Aykut-Bingol et al., 1998; Bautista et al., 1999; Boesebeck et al., 2002; Barba et al., 2005; Dalmagro et al., 2005). Previous studies have correlated younger age at onset (Blume et al., 1991), and longer epilepsy duration (Dalmagro et al., 2005) with worse outcomes. Age at surgery was only significant in our cohort on univariate analysis. Analyzing this issue further, we found that 67% of our patients who were younger than 18 years had epilepsy secondary to tumors or lesional MCD as opposed to only 29% of those who were older (p = 0.009). This suggests that age at surgery may be a surrogate marker of etiology—rather than an independent outcome predictor—with younger patients being more likely to have tumors or lesional MCD, a group with the most favorable outcome.

A previous study found that more than half the patients with either a lateralizing aura or lateralizing seizure were seizure-free after PCE surgery as opposed to 17–28% of patients with no lateralizing semiologic findings (Boesebeck et al., 2002). But our results are in concordance with other studies that failed to show a prognostic significance of lateralizing semiology (Blume et al., 1991; Dalmagro et al., 2005).

The negative prognostic implications of a history of generalized tonic–clonic seizures and immediate postoperative seizures identified with temporal and frontal lobe epilepsy surgery (McIntosh et al., 2004; Jeha et al., 2006, 2007) were not reproduced here.


Neurologic deficits occurred at a rate comparable to that previously described elsewhere (Blume et al., 1991; Dalmagro et al., 2005; Kun Lee et al., 2005), with complications extending beyond de novo visual field defects in some.

None of the four patients with language deficits had Wernicke’s area mapped preoperatively, but none had resections involving their posterior temporal lobes either. So, although postoperative changes or transient edema leading to mass effect on the posterior language area may underlie those deficits, our results do not necessarily support that hypothesis. On the other hand, all three aphasic cases had at least part of their dominant supramarginal gyrus and/or angular gyrus resected, raising concerns for language deficits occurring secondary to injury to the inferior parietal lobule sparing Wernicke’s. The alexia without agraphia probably occurred from transient postoperative changes affecting the posterior extent of the corpus callosum.


This is the first longitudinal study reporting on seizure outcome following PCE surgery in a well-characterized subset of patients with medically intractable epilepsy at a tertiary epilepsy surgery center. We show the following:

  • 1The long-term chances of SF are promising, and surgical failures can be identified early postoperatively.
  • 2More aggressive resections in the posterior cortex optimize chances of SF, although lesionectomies may be an acceptable option in patients with tumors or high-grade MCD.
  • 3A careful evaluation of the epileptogenicity of the ipsilateral temporal lobe and need for resection is warranted in the presence of ipsilateral preoperative temporal spiking or imaging abnormalities.
  • 4Routine 6 month postoperative EEG may provide useful prognostic information for both the recurrence and severity of postoperative seizures.
  • 5Language deficits may occur with resections involving the dominant inferior parietal lobule and sparing Wernicke’s area.
  • 6Further research is needed to evaluate the issue of AED discontinuation following surgery.


We are grateful to the support of our colleagues at the Cleveland Clinic Epilepsy Center and to our patients. No funding source contributed to this project.

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.

Disclosure: None of the authors has any conflict of interest to disclose.