Seizure worsening and its predictors after epilepsy surgery

Authors


Address correspondence to Lara Jehi, Neurological Institute, Cleveland Clinic, 9500 Euclid Ave, S-51, Cleveland, 44195 OH, U.S.A. E-mail: jehil@ccf.org

Summary

Purpose:  This study aims to investigate seizure worsening and its predictors after epilepsy surgery.

Methods:  A retrospective chart review of patients who underwent unilobar epilepsy surgery between 1990 and 2007 and had recurrence of at least one seizure was performed. Seizure worsening was defined as an increase in total average monthly seizure frequency, average monthly generalized tonic–clonic seizures (GTCS), new-onset GTCS, or new-onset status epilepticus. The occurrence of sudden unexpected death in epilepsy (SUDEP) was captured. Multivariate logistic regression analysis was used to identify predictors of worsening.

Key Findings:  A total of 276 patients with postoperative seizure recurrence were identified. Monthly average seizure frequency worsening occurred in 9.8%, GTC worsening in 8.0%, new-onset GTCs in 1.4%, new-onset status epilepticus in 2.2%, and death from SUDEP in 1.4%. A higher risk of worsening was seen with extratemporal resections as compared to temporal lobe surgeries (odds ratio [OR] 3.11, 95% confidence interval [CI] 1.21–7.95; p = 0.018), and in patients with low preoperative seizure frequency <30 seizures/month (OR 14.82, 95% CI 2.81–275.41; p = 0.0003). Predictors of increased GTCs included an incomplete resection (OR 3.98, 95% CI 1.39–12.59; p = 0.010) and multiple recorded ictal patterns (OR 5.91, 95% CI 1.20–26.96; p = 0.030). Multiple seizure semiologies correlated with worsening after temporal lobe resections.

Significance:  The most vulnerable patients for seizure worsening following epilepsy surgery include those with extratemporal resections, incomplete resections, and multiple recorded ictal patterns.

Complete seizure freedom is the ultimate aim of epilepsy surgery for the treatment of medically refractory epilepsy. However, some patients experience occasional postoperative seizures only to eventually become seizure-free, a phenomenon termed “running down” (Salanova et al., 1996). Others never become seizure-free again after a recurrence, but do at least experience a sustained improvement in seizure frequency. Unfortunately, others do not benefit from surgery at all, and their seizures may even worsen postoperatively.

Despite extensive literature discussing long-term rates of seizure freedom after surgery and their predictors, little is known on how often the phenomenon of seizure worsening occurs after surgery, and its predictors. Using their own seizure scoring system, Ficker et al. (1999) showed that there was deterioration in seizure control in 11% of their temporal lobectomy and 5.1% of their frontal lobectomy patients. Epilepsy surgery outcome studies do not uniformly report on seizure worsening, although the Engel classification (1987) includes a class IV C dedicated to describing patients with worsening seizures, most series lump all their class IV outcomes together thus rendering it difficult to know the percentage of patients with worsening. More recent studies using the International League Against Epilepsy (ILAE) classification of surgical seizure outcome (Wieser et al., 2001) report 0.8–8.4% rates (von Lehe et al., 2006; Ozkara et al., 2008; Stavrou et al., 2008; Chang et al., 2011) of ILAE category 6 (>100% increase in baseline seizure days). These studies restrict their assessment of seizure worsening to an increase in seizure frequency, whereas the phenomenon of worsening seizure severity through the appearance of new onset generalized tonic–clonic seizures (GTCS) after anterior temporal lobectomy (Blume & Girvin, 1997) remains largely unaccounted for.

Of note, the few studies evaluating the natural history of intractable medically treated epilepsy also usually dichotomize their analyses into “seizure-free, or remission” versus “not seizure-free,” losing the nuance of who is actually worsening over time (Callaghan et al., 2011; Choi et al., 2011), further confirming the need to explore this question.

In this report studying patients with at least one recurrent postoperative seizure, and comparing those who worsened to those who subsequently improved or did not have a change in frequency, we aim to: (1) investigate various aspects of postoperative seizure worsening, including the manifestation of new-onset GTCS or status epilepticus, (2) identify factors correlated with such worsening, and (3) evaluate mortality in our cohort (including that related to sudden unexpected death in epilepsy [SUDEP]). Such a comparison was chosen, rather than one opposing worsening patients to those completely seizure-free after surgery, in order to avoid as much as possible defaulting to the identification of general predictors of postoperative seizure freedom, and focus on those more tightly related to seizure worsening per se.

Methods

Patient selection

We retrospectively reviewed records of patients who underwent unilobar epilepsy surgery at the Cleveland Clinic Epilepsy Center between 1990 and 2007 to identify those with at least one postoperative seizure, and subsequent follow-up of at least 6 months. Patients with progressive neoplastic or neurodegenerative diseases were excluded. Based on these criteria, and after a review of 914 charts, a total of 276 patients fulfilling study criteria were identified. Details of this cohort have previously been well described (Jehi et al., 2010). Average monthly seizure frequency was noted based on the patient or family member’s report. Patients who were undergoing AED withdrawals or tapers at last follow-up had their seizure frequency prior to the changes included as their seizure frequency at last follow-up. The study was approved by the Cleveland Clinic Institutional Review Board.

Preoperative characterization

All patients received an initial clinical evaluation in the outpatient setting, a prolonged video-scalp electroencephalography (EEG) (Nihon Kohden EEG 9200, Nihon Kohden EEG 1100, and Nihon Kohden EEG 1200, Nihon Kohden, Tokyo, Japan) evaluation and a high-resolution brain magnetic resonance imaging (MRI) was performed with a Siemens 1.5 or 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 sequences. A fluorodeoxyglucose (FDG) positron emission tomography (PET) scan (Biograph MCT, Siemens) was performed when deemed necessary. For the purpose of this study, the MRI and PET scans were classified as: (1) normal, (2) unilaterally abnormal, (3) bilaterally abnormal.

The results of the preoperative workup were then discussed at a multidisciplinary patient management conference, and a recommendation was made to proceed with resective surgery, withhold such intervention, or perform further testing that may include an invasive EEG evaluation with or without other such as ictal SPECT (Sybia T16, Siemens), Wada test, or functional MRI.

Surgery and pathology

Patients were divided into temporal (73%) and extratemporal (27%) groups based on the site of surgical resection. Surgeries performed included lesionectomy, lobectomy, selective amygdalohippocampectomy, and tailored neocortical resections based on subdural electrode (SDE) with or without depth electrode recordings. For lesional cases, surgical resections were considered complete if there was complete removal of the MRI lesion on postoperative MRI. 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 electrocorticography (ECoG) evaluations. Neuropathologic results were grouped into the following five categories: (1) malformations of cortical development (MCDs), (2) tumors, (3) vascular abnormalities, (4) gliosis, and (5) hippocampal sclerosis (HS).

Postoperative follow-up

All patients were seen at 6 weeks, 6 months, 1 year, and then yearly as part of the Cleveland Clinic Epilepsy Center standard follow-up protocol. If the patient had seizure recurrence, then more frequent follow-up visits were scheduled. Six months after surgery, all patients had outpatient scalp EEG studies. Antiepileptic drug levels were routinely assessed, especially after an increase in seizure frequency.

If a patient underwent a second surgery or insertion of a vagus nerve stimulator (VNS) then for the purposes of statistical analyses performed here, the date of last follow-up was the date of second surgery or VNS implantation.

Definitions

Seizure worsening

At last follow-up, the patients’ seizures were considered to worsen if (1) their average monthly total seizure frequency was 100% greater than their presurgical baseline placing them in the ILAE class 6 category (the comparison group consisted of patients with postoperative improvement or no change in total seizure frequency), or (2) they had a >100% increase in their presurgical GTCS frequency (the comparison group consisted of patients with preoperative GTCS and postoperative improvement or no change in frequency of GTCS), and/or (3) had new onset status epilepticus (a single clinical seizure lasting >30 min or repeated seizures over a period of >30 min without intervening recovery of consciousness) (Knake et al., 2009) or new-onset GTCS.

Patients who had seizure worsening upon antiepileptic drug (AED) withdrawal but returned to their baseline seizure frequency were not included in the seizure-worsening group.

SUDEP determination

A social security death index query was performed to ascertain patient deaths during the follow-up period. Death certificates were then requested on all, as well as autopsy reports, when possible. The modified Nashef criteria (Nashef & Brown, 1996) were used to ascertain SUDEP.

Statistical analyses

Data were summarized using descriptive statistics for both continuous and categorical variables. Univariate analysis was used using student’s t-test, and chi-square and Fisher exact tests to compare patients with seizure worsening to those with no worsening.

A subgroup analysis of predictors of worsening in both temporal and extratemporal groups was also performed. A similar approach was used to determine predictors of worsening of GTCS frequency. Potentially significant variables (significance level of 10% on univariate analysis) were then tested in a multivariate logistic regression analysis model (significance level of 5%).

Results

Of 276 patients analyzed and followed for an average of 6.57 years (±SD 4.29), 27 (9.8%) had a worsening of their baseline seizure frequency, 18 (8.0%) patients had worsening of their GTCS frequency, four (1.4%) had new onset GTCS, and six (2.2%) had new onset status epilepticus. Overall 51 patients (18.5%) had at least one of these outcomes, of whom four had various combinations of the above. Alternatively, 64 (23%) of the patients benefited little from surgery and were classified as ILAE category 5 (<50% reduction up to 100% increase). The rates for the overall cohort of the 914 surgical cases are 3.0% for seizure worsening, 2.0% for worsening GTCS, 0.4% new-onset GTCS, and 0.7% for new-onset status epilepticus.

The characteristics of the patients whose seizure frequency worsened and those who did not exhibit seizure worsening postoperatively are reported in Table 1. The average baseline seizure frequency for those who worsened was 8.4 seizures per month (range 0.33–60), whereas those who did not worsen had an average frequency of 33.6 (range 0.1–600) (p = 0.065). Of the patients who worsened, seizure frequency had at least doubled in 11 (41%) by the first postoperative year. Similarly, when the criterion for worsening was new-onset GTCS or status epilepticus, these occurred within the first postoperative year in 30% of the patients.

Table 1.   Characteristic of the cohort
 No worsening (n = 249)Worsening (n = 27)p-Value
  1. The “no worsening” group consisted of patients with either improvement or no change in total seizure frequency.

  2. ECoG, electrocorticography; HS, hippocampal sclerosis; MCD, malformation of cortical development; SW, sharp wave.

  3. *p < 0.10.

Clinical and seizure characteristics   
 Male127 (51%)14 (52%)0.93
 Age at surgery (in years ± SD)31.9 ± 12.930.6 ± 16.30.63
 Age at seizure onset (in years ± SD)14.7 ± 13.316.6 ± 17.80.48
 Preoperative GTCS205 (83%)21 (78%)0.52
 Multiple seizure semiologies26 (10%)5 (19%)0.18
 Preoperative aura193 (78%)20 (74%)0.66
 ≥30 preoperative seizures/month69 (27%)2 (7%)0.02*
Imaging (%)   
 MRI   
  Normal41 (16)5 (19)0.20
  Abnormal unilateral187 (75)17 (62)
  Abnormal bilateral21 (8)5 (19)
 PET   
  Normal15 (8)2 (9)0.83
  Abnormal unilateral145 (82)20 (87)
  Abnormal bilateral17 (10)1 (4)
Preoperative scalp EEG (%)   
 Multiple ictal patterns15 (6)00.19
 Nonlocalizable/generalized pattern17 (7)2 (7)0.84
 Contralateral temporal SW47 (19)4 (15)0.61
 Contralateral extratemporal SW2 (1)00.64
 Contralateral temporal seizure pattern8 (3)2 (7)0.27
 Contralateral extratemporal seizure pattern2 (1)00.64
Surgery (%)   
 Extratemporal resection60 (24)13 (48)0.01*
 Incomplete resection89 (36)13 (48)0.20
 Left-sided139 (56)15 (56)0.98
 Use of subdural electrodes85 (34)14 (52)0.07*
 Use of ECoG55 (24)7 (33)0.34
 Presence of ipsilateral spike on postoperative EEG99 (42)16 (67)0.02*
 Pathology   
  Gliosis70 (28)8 (30)0.54
  MCD45 (18)8 (30)
  HS85 (34)8 (30)
  Tumor42 (17)3 (10)
  Vascular7 (3)0

Using univariate analysis, the following variables were significant predictors of worsening: preoperative seizure frequency, use of subdural electrodes, ipsilateral spikes on the postoperative EEG, and the type of surgery. With multivariate analysis, only preoperative seizure frequency and type of surgery remained as significant predictors of postoperative seizure worsening (Table 1). A higher risk of worsening was seen with extratemporal resections as compared to temporal lobe surgeries (OR 3.11, 95% CI 1.21–7.95; p = 0.018). In addition, patients with low preoperative seizure frequency <30 seizures/month (OR 14.82, 95% CI 2.81–275.41; p = 0.0003) also exhibited seizure worsening. Of the patients who worsened, 11 underwent a reoperation resulting in seizure freedom in 5, seizure improvement in 4, and no change in seizure control in 2. On the other hand, four patients had VNS placement and only one had some improvement in seizure control.

A separate analysis using the ILAE class 5 patients as controls was also performed (see Table S3); however, the numbers were not sufficient to infer robust predictors of worsening.

A subgroup analysis of predictors of worsening in the different surgical groups (temporal vs. extratemporal) was then performed (Tables 2, S1, and S2). The preoperative seizure frequency was the only variable that predicted worsening in the extratemporal group. In the temporal group, and after multivariate analysis, the documentation of multiple seizure semiologies (OR 5.74, 95% CI 1.53–20.06; p = 0.007) retained predictive significance. The most common combination was automotor and dialeptic (absence with paucity of motor movements and unresponsiveness during ictal testing) seizures in 5 patients, dialeptic and generalized seizures in 4, automotor and generalized in 3, automotor and versive in 2, and others in 7. The dialeptic and automotor seizures were felt to be distinct due to the presence of different preceding auras, or prominent automatisms (automotor), among other features. Of the patients who worsened, three had a repeat video-EEG monitoring admission. All three patients had known seizures from both the left and right temporal lobes; the “intact” temporal lobe was found to be the source of the worsening seizure frequency postoperatively.

Table 2.   Predictors of seizure worsening based on surgery type
Extratemporal
 No worsening (n = 60)Worsening
(n = 13)
p-Value
Preoperative seizure frequency/month (%)
 <30 seizures/month31 (52)11 (85)0.029
 ≥30 seizure/month29 (48)2 (15)
Temporal
 No worsening (n = 189)Worsening
(n = 14)
p-Value
Use of subdural electrodes (%)
 Yes49 (26)7 (50)0.052
 No140 (74)7 (50) 
Multiple seizure semiologies (%)
 Yes16 (8)5 (36)0.0012
 No173 (92)9 (64)
Preoperative seizure frequency/month (%)
 <30 seizures/month149 (79)140.054
 ≥30 seizure/month40 (21)0

Worsening of GTCS at last follow-up

Eighteen of the 226 patients who reported preoperative GTCS had worsening of their seizures at last postoperative follow-up. Table 3, summarizes the variables identified as significantly correlating with worsening upon univariate analysis. With multivariate analysis, only an incomplete resection (OR 3.98, 95% CI 1.39–12.59; p = 0.010) and multiple recorded ictal patterns (OR 5.91, 95% CI 1.20–26.96; p = 0.030) retained significance.

Table 3.   Predictors of worsening of generalized tonic–clonic seizures (univariate analysis)
 No worsening (n = 208)Worsening (n = 18)p-Value
  1. The “no worsening” group consisted of patients with preoperative GTCS and postoperative improvement or no change in frequency of GTCS.

Incomplete resection (%)
 Yes68 (33)12 (67)0.0038
 No140 (67)6 (33)
Nonlocalizable/generalized pattern (%)
 Yes10 (5)5 (27)0.0002
 No198 (95)13 (73)
Multiple ictal patterns (%)
 Yes7 (3)5 (27)<0.0001
 No201 (97)13 (73)

New-onset GTCS

Of the patients with seizure recurrence after epilepsy surgery, four developed new-onset GTCS at a mean postoperative duration of 6.2 ± SD 5.7 years.

Three of the patients had a temporal lobectomy, whereas one patient had a posterior quadrant resection. Pathology had revealed hippocampal sclerosis in two and tumor in two patients. One patient had multiple seizure semiologies and none had multiple ictal patterns. The effect of these variables on the occurrence of new-onset GTCS is difficult to determine due to the small numbers.

New-onset status epilepticus

Six patients developed new-onset status epilepticus after surgery; three of them within the first 3 months after surgery, whereas the rest occurred afterward (range 1–6 years). Two of the patients had status epilepticus in the immediate postoperative period (within 1 day of surgery); one of the patients had a rolandic Taylor-type cortical dysplasia (type 2B dysplasia), whereas the second had a temporal lobectomy with the reported pathology showing only gliosis.

One patient had a posterior quadrant neoplasm (low-grade fibrillary astrocytoma) resected, and the rest had hippocampal sclerosis on pathology. Only one patient had multiple semiologies, and none had multiple ictal patterns. Only one was provoked by noncompliance with AEDs. The duration of epilepsy had been 21.0 ± SD 8.9 years at the time of surgery in these six cases.

Death

Eight patients died during the follow-up period. None belonged to the seizure-worsening group, yet all had persistent seizures after surgery (Engel class III–IV, Wieser classes 3–5). Causes of death included SUDEP (confirmed following autopsy in one, and probable in three), whereas one patient died from lung cancer and another patient died from a seizure-related fall. The cause of death remained unknown in the two remaining cases.

Discussion

This is the first study attempting to systematically address the issue of seizure worsening after epilepsy surgery, evaluating the frequency of its occurrence and identifying its possible predictors. In patients who had any postoperative recurrence, we identified a rate of seizure frequency worsening (increase) of 9.8%, GTCS frequency worsening of 8.0%, new-onset GTCS of 1.4%, new-onset status epilepticus of 2.2%, and death from SUDEP of 1.4%. Changing the denominator from the 276 patients who had any seizure recurrence to the overall cohort of 914 surgical cases, the risks are obviously lower: 3.0% for seizure worsening, 0.4% for new-onset status epilepticus, and 0.7% for SUDEP. Overall, our identified rate of increased seizure frequency is then consistent with the limited available data from prior epilepsy surgery studies reporting rates of worsening whether Engel class IVC or ILAE class 6. A 9.6% risk of increased seizure frequency was described in rolandic epilepsy (Pondal-Sordo et al., 2006), 8.4% in patients with malformations of cortical development (Chang et al., 2011), 3% in patients with hippocampal sclerosis (Ozkara et al., 2008), and 2.8% in patients with cavernomas (Stavrou et al., 2008). No definite rates of new-onset GTCS or status epilepticus were previously published. As for mortality and SUDEP risk, our results are a further confirmation that patients who have had failed surgery are at highest risk (Sperling et al., 1999).

However, the significance of these numbers requires careful consideration. By definition, epilepsy is a dynamic condition with periods of frequent and severe seizures intermixed with periods of remission (Kwan & Sander, 2004). So one needs to be careful while analyzing how a drastic procedure—such as resective epilepsy surgery—interfered with this variable process in our cohort of “worsening” patients. The following two possibilities arise. First, surgery may have expedited the course of epilepsy progression, leading to more severe and frequent seizures. This is hypothetically possible if, for example, a temporal lobectomy is performed in a “pseudo-temporal epilepsy” patient, disrupting the ictal spread patterns originally leading to a more benign dialeptic/complex partial semiology, and favoring suprasylvian spread to primary motor cortices and therefore facilitating secondary generalization. This is supported in our data by the early postoperative timing of new-onset GTCS and status epilepticus in some patients with long-standing refractory epilepsy. Second, alternatively, surgery may have slowed down the progression of intractable epilepsy in the cohort as a whole, such that these rates would have been higher had we not intervened by performing a resection. This may be supported by data from the placebo arm of a few AED trials, which showed that at some point almost 50% of epilepsy patients experienced an increase in their baseline seizure frequency (Somerville, 2002), far higher than the rates seen in the current study.

Similarly, Carreno et al. (2011) showed a rate of worsening of 16.3% in their cohort of 80 patients with medically refractory epilepsy who did not undergo surgery.

An appropriate answer to this question requires a careful, comprehensive, prospective study of the natural history of epilepsy with head-to-head comparisons of surgery and medical therapy.

Predictors of worsening

Regardless of whether postoperative seizure worsening is due to surgery itself or to the natural history of the disease, an identification of the patients at risk for worsening should help in the risk/benefit calculation and evaluation of a patient’s candidacy for resective epilepsy surgery.

Type of surgery

Our results show higher rates of seizure worsening in the extratemporal surgery population. Extratemporal resections tend to have poorer surgical outcomes than their temporal counterparts due to multiple anatomic and localization issues resulting in erroneous localization/less accurate delineation of the epileptic focus and/or the involvement of eloquent cortex (Roper, 2009). Previous studies showed a higher occurrence of focal cortical dysplasia (FCD) in the extratemporal distribution (Palmini, 2000). We previously showed that types I and IIA FCD may harbor overlapping function and intrinsic epileptogenicity, and therefore lead to difficulties in “complete” removal of the epileptic focus due to concerns about creating functional deficits (Marusic et al., 2002; Sarkis et al., 2010). In addition, we previously reported that a failure of a first resection to control seizures in a small number of patients with rolandic epilepsies due to FCD could be followed by seizure freedom should a more aggressive resection be performed (with postoperative deficits) (Sarkis et al., 2010). Type IIB FCDs (containing balloon cells) show an inverse correlation between pathology severity and functional/epileptic expression, whereas the subareas that are rich in balloon cells are less functional and less epileptic and the typically surrounding dysplastic cortex that is devoid of balloon cells is epileptogenic and may be functional. In these patients, the resection of the “lesion” with an aim toward preserving function may not lead to seizure control. The lack of intrinsic epileptogenicity in the balloon cell–rich regions may be attributed to a potential antiepileptic role of balloon cells. We recently showed that balloon cells may possess glutamate clearance mechanisms through the high expression of glial glutamate transporter proteins in these cells (Gonzalez-Martinez et al., 2011). In these cases, we hypothesized that a resection of the balloon cell–rich area but not the surrounding dysplastic (and highly epileptic) would potentially disrupt glutamate clearance mechanisms and may therefore worsen seizures postoperatively. On the other hand, another study did not find a statistically significant difference between temporal and frontal patients in terms of a decline in seizure control (Ficker et al., 1999). However, the findings in this study are difficult to compare to our observations as the authors used their own seizure scoring system and did not include patients who underwent posterior quadrant resections.

Seizure frequency

We found that patients with <30 seizures per month tended to have higher rates of worsening. This could be due to a reporting factor due to the qualitative nature of the reported frequency, as patients with frequent seizures may notice less of an increase in their frequency; the other possibility is that certain epileptogenic networks may have a ceiling effect and no room to further increase seizure frequency. Further studies with larger number of patients are needed to further validate these observations.

Multiple semiologies

In the subgroup analysis, the occurrence of multiple seizure semiologies was the only predictor of worsening in patients undergoing a temporal lobectomy. Multiple semiologies theoretically represent different seizure foci and are likely to be marker of a worse postoperative outcome (Abou-Khalil, 2008). The resection itself may allow the facilitation of expression of one focus, assuming one of the foci is removed or the epilepsy itself may have originated extratemporally and now has new pathways to express itself. This certainly remains speculative and will require further research: a careful comparison of the postoperative and preoperative seizure semiology and ictal patterns would help to further validate this hypothesis. In our retrospective cohort, only three of patients with postoperative seizure worsening underwent a repeat video-EEG evaluation. In these cases, the semiology and ictal patterns were felt to be arising from the unresected known epileptogenic temporal lobe. Furthermore, in our cohort, none of the vascular and few neoplastic lesion surgeries (10%) worsened. The others were of almost equal percentages (HS, gliosis, MCD), suggesting a stronger implication of multiple semiologies in these time-dependent dynamic epileptic substrates.

Worsening of new-onset GTCS

Worsening of the frequency of GTCs occurred in 8.0% of our patients. This phenomenon was noted by Henry et al. (2000) in their follow-up of 60 patients who underwent anterior temporal lobectomies. They identified six patients who had such an increase, and postulated that a residual/nonresected epileptogenic tissue had a higher chance of propagating to brainstem and subcortical structures leading to an increased propensity toward generalization. Our results are consistent with these reports, as we found that an incomplete resection was a likely predictor of an increase of postoperative GTCS. In addition, our data suggest that this phenomenon is not only restricted to temporal lobe resections but it is also observed in patients who underwent extratemporal resections. Another predictor of such an increase was the presence of multiple ictal patterns recorded preoperatively. This observation may suggest the presence of multiple epileptic foci, or alternatively may be an indicator of the presence of a more complex epileptogenic network with a higher postsurgical tendency to spread to subcortical structures.

AED withdrawal was reported to be the cause of new-onset GTCS after temporal lobectomy in a study by Blume and Girvin (1997). We documented new-onset GTCS in four patients from our cohort, with AED noncompliance being a possible cause of GTCS in only one patient. The occurrence of de novo GTCS following AED withdrawal may be due to a release of larger epileptogenic networks, therefore resulting in faster generalization of the epileptic activities leading to a worsening in clinically observed GTCS.

New-onset status epilepticus

Our study shows that status epilepticus may occur at any time following epilepsy surgery, although half of the occurrences happened during the first six postoperative months. The occurrence of status epilepticus following epilepsy surgery is a rare event. Burneo et al. (2005) described a patient with nonconvulsive status epilepticus following a right temporal lobectomy and attributed this phenomenon to brain injury due to the surgery itself. We have previously described three cases of rolandic focal cortical dysplasia (one of them included in the current series) with balloon cells, having status epilepticus during the immediate postoperative period (Sarkis et al., 2010). We previously attributed the immediate postoperative status epilepticus in our patients with rolandic epilepsy to a possible “release of inhibition” of an epileptogenic network following surgery. However, the occurrence of late-onset status epilepticus years after surgery is most probably due to other mechanisms that remain largely unknown (including a possible progression of the epilepsy). Around 15–20% of patients with epilepsy will have status epilepticus at least once (Tatum et al., 2001). The most commonly cited causes of status epilepticus in adults include cerebrovascular disease and AED withdrawal (Neligan & Shorvon, 2010), although AED withdrawal in itself may not simply be the direct cause, and remote factors (such as the nature of the initial insult) or acute factors may be playing a role (Barry & Hauser, 1993). The fact remains that animal models have provided numerous pathophysiologic explanations for why status occurs, but well-designed studies in humans are still lacking (Murdoch, 2007).

Study Limitations

One of the limitations of the study is its retrospective nature, and absence of systematic determination of AED levels. In addition, due to the qualitative nature of seizure reporting and lack of diaries, an element of recall bias cannot be ruled out.

Identifying an appropriate control group is also challenging; a prospective design comparing the natural history of refractory epilepsy after surgery as opposed to that of medically treated patients with intractable seizures would be ideal. Overall, the numbers of the patients who worsened in our cohort was also limited; as a result, subgroup analyses should be interpreted with caution.

Conclusion

The most vulnerable patients for seizure worsening following epilepsy surgery include those with extratemporal resections, multiple ictal patterns, or those who undergo an incomplete resection of their epileptic focus. Future studies will be required to elucidate the pathophysiologic processes contributing to this phenomenon.

Disclosure

Dr Najm serves on the speaker’s bureau for UCB Pharma. The remaining authors have no conflicts of interest. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Ancillary