- Top of page
Summary: Purpose: Surgery for medically intractable epilepsy is currently the most effective means of achieving seizure control. Although relatively few long-term outcome studies have been performed, evidence is mounting that the possibility of late seizure recurrence exists, even after an early seizure-free period. No published reports document the rate and predictors of late recurrence in a large series of patients undergoing surgery in the magnetic resonance imaging (MRI) era.
Methods: We retrospectively queried the databases of two epilepsy surgery centers. Patients eligible for study had preoperative MRI scans, were seizure free for 1 year after surgery, and had a minimal follow-up period of 3 years. Patients with tumors or vascular lesions were excluded. We performed log-rank comparison of Kaplan–Meier product limit estimates for categoric variables and used a Cox proportional hazards model for continuous variables. Variables that were significant (p < 0.05) on a univariate screen were entered into a multivariate forward step-wise Cox regression.
Results: The study included 285 patients, 254 with medial temporal lobe (MTLE) and 31 with neocortical epilepsy. The probability of having a single seizure after being seizure free for 1 year was 18.3% at 5 years and 32.7% at 10 years. However, only 13% were not seizure free at the last follow-up. Predictors of late recurrences on both uni- and multivariate analysis were the presence of preoperative generalized tonic–clonic (GTC) seizures in patients with neocortical epilepsy and late age at surgery in patients with MTLE. MRI results and location of surgery were not predictive.
Conclusions: Although the risk of at least one recurrent seizure after initially successful epilepsy surgery is relatively high, the rate of recurrent intractability is low. The finding that late age at surgery and presence of preoperative GTC seizures are predictors of late recurrence indicates the importance of patient selection and early surgery for persistent seizure control.
Resective brain surgery is an extremely effective treatment for medically intractable epilepsy (1–3). The results of numerous short- and long-term longitudinal studies have demonstrated persistently high rates of seizure control after mesial temporal resections and stable, but lower rates of control after neocortical resections (1–26). Recently, however, studies have proposed that the rate of late recurrence may be higher than previously reported, with seizures reappearing even as late as 10 years after surgery in patients who were initially seizure free (27,28). Some of these late seizures can be precipitated by medication withdrawal and do not necessarily develop into recurrent seizures (27,29). In addition, there exists a group of patients who continue to have a few seizures soon after surgery and then become seizure free with longer follow-up, or “run down over time” (30). It is not clear whether these late recurrences are evenly balanced by the late successes or whether the overall trend is toward decreasing seizure control over time.
In prior studies of long-term outcome in patients who are initially seizure free, attempts were made to correlate certain factors with late recurrence, such as history of generalized tonic–clonic (GTC) seizures (24,28), normal magnetic resonance imaging (MRI) scans (28), normal histology (27), older age at surgery (24,27,28), and longer history of seizures (24,27). However, these were single-institution studies that required several years to accumulate adequate numbers of patients. Hence, imaging for many patients was performed in the pre-MRI era (27,28), and the studies included lesions, such as vascular malformations and tumors, that are known to have a better postoperative course (27,28). Additionally, many of these studies excluded neocortical epilepsy and examined outcome only after surgery for medial temporal lobe epilepsy (MTLE), thereby limiting the scope of their conclusions (24,28). For this reason, we assembled a large population of patients from two separate epilepsy centers that had surgery for either medial temporal lobe or neocortical epilepsy and examined the incidence and predictors of late recurrence. All patients had preoperative MRI scans and were seizure free for ≥1 year after surgery.
- Top of page
We retrospectively queried the databases of two epilepsy surgery centers, Center A (Thomas Jefferson University) and Center B (Cleveland Clinic Foundation). Patients included in our study all had medically intractable focal epilepsy and underwent potentially curative resective brain surgery after review in a multidisciplinary epilepsy surgery conference. For this study, the presence of auras was permitted in the category of seizure free. All patients were seizure free for ≥1 year after surgery and had a minimum of 3 years of postoperative follow-up. All patients had preoperative MRI, video-EEG, and Wada testing. The histologic analysis of resected tissue was available for all patients. Surgical technique varied between centers. Patients with neocortical tumors or vascular malformations were excluded. On the basis of their preresection data including EEG, electrocorticography (ECoG), and MRI results, patients were categorized into one of four etiology groups as follows: (a) MTLE with atrophy, (b) MTLE with normal MRI, (c) neocortical onset with abnormal MRI, and (d) neocortical onset with normal MRI. Histology was not included in this grouping because it was not known before surgery. For group 1, “atrophy” was based on the impression of a board-certified neuroradiologist with experience in epilepsy imaging as either medial temporal lobe atrophy or atrophy of the entire temporal lobe. Volumetric measurements were not consistently performed. For group 3, “abnormal” MRI consisted of the neuroradiologist's impression of gliosis, encephalomalacia, or cortical dysplasia.
The following data were included in our statistical analysis: age at seizure onset, years with seizures, preoperative seizure frequency, presence of preoperative GTC seizures, presence of preoperative auras, age at surgery, side of surgery, positron emission tomography (PET) results (normal, abnormal, not done), use of subdural electrodes or intraoperative ECoG, pathology of resected tissue [normal, mesial temporal sclerosis (MTS), gliosis, cortical dysplasia, or cortical dysplasia plus MTS], and precipitating factors for seizure recurrence (none, medication reduction, stress, lack of sleep, or fever).
We performed log-rank comparison of Kaplan–Meier product limit estimates for categoric variables and used a Cox proportional hazards model for continuous variables. The end point was time of first seizure after the first year of being seizure free. Continuous variables also were stratified by terciles and then reanalyzed by log-rank tests. Variables that were significant (p < 0.05) on a univariate screen were entered into a multivariate forward step-wise Cox regression. In addition, before the results from Center A and Center B were grouped, we examined the impact of each center on etiology, seizure recurrence, age at seizure onset, age at surgery, years with seizures, and preoperative seizure frequency. P values <0.05 were considered statistically significant.
- Top of page
In total, 285 patients were included in this study, 162 patients from Center A and 123 from Center B. The only significant differences between centers were age at seizure onset and years with seizures. Patients from Center B had a younger age at seizure onset (p = 0.006) and a longer time with seizures (p = 0.005). However, because these factors were not found to be related to seizure recurrence, the center where the surgery was performed was not a confounding variable, and results from both centers were grouped for all subsequent analyses.
Of the patients, 254 had MTLE, and 31 had neocortical epilepsy. The mean (± SD) age at surgery was 33.4 ± 9.6 years, mean (± SD) age at seizure onset was 12.4 ± 10.3 years, mean (± SD) preoperative seizure frequency (per month) was 15.6 ± 31.9, and the mean (± SD) years with seizures was 21.0 ± 10.7. The mean duration of follow-up was 7.9 (± 3.1) years. Auras were present preoperatively in 235 patients, and generalized tonic–clinic seizures were present in 154 patients. Surgery was on the left hemisphere in 137 patients. PET scans were obtained in 159 patients, of which 144 were read as abnormal. Subdural electrode recordings were obtained in 106 patients. Clinical data are presented for each subgroup in Table 1. The histology of resected tissue is presented in Table 2.
Table 1. Clinical characteristics by etiology subgroup
|MRI||Medial temporal lobe||Neocortical|
|Auras||75 (78)||133 (84) ||10 (91)||17 (85)|
|GTC||54 (56)||76 (48)||10 (91)||14 (70)|
|Left||42 (44)||82 (52)|| 3 (27)||10 (50)|
|Preop seizure frequency/moa||13.1 ± 17.4||15.9 ± 32.9|| 8 ± 4.3||35.6 ± 73.7|
|Years with seizuresa||18.9 ± 9.1 || 23 ± 11.1||19.3 ± 10.2||23.3 ± 12.9|
|Age at seizure onset (yr)a||14.1 ± 10 ||10.5 ± 9.9 ||10.9 ± 10.5||12.7 ± 12.5|
|Age at surgery (yr)a||33 ± 8.5||33.5 ± 10.1||30.3 ± 10.6||36 ± 10|
|PET||48 (50)||88 (57)||10 (91)||13 (65)|
|Abnormal||43 (90)||81 (92)|| 8 (80)||12 (92)|
|Sub electrodes||49 (51)||48 (30)|| 4 (36)|| 5 (25)|
|Mean follow-up (yr)||6.6||5.3||4.9||5.3|
Table 2. Histology of resected tissue
|MRI||Medial temporal lobe||Neocortical|
|MTS||56 (58)||111 (70) ||0||2 (10)|
|Gliosis||6 (6)||18 (11)||6 (54)||6 (30)|
|Normal||27 (28)||16 (10)||0||2 (10)|
|Dysplasia||6 (6)||4 (3)||4 (4)||10 (50) |
|Dysplasia + MTS||1 (1)||9 (6)||1 (1)||0|
After being seizure free for 1 year after surgery, at the last follow-up, 28% of patients had a subsequent seizure. Although this number was slightly higher in patients with neocortical onsets (32%) and in patients with normal MRI scans (33%), the difference did not achieve statistical significance (Table 2). With Kaplan–Meier survival analysis, the probability of having a seizure after 1 year of being seizure free was 18.3% at 5 years and 32.7% at 10 years. For each subgroup, the actual incidence was as follows: group 1, 19.6%; group 2, 19.9%; group 3, 15%; and group 4, 29.9%. The 10-year recurrence-free survival was as follows: group 1, 29.8%; group 2, 35.8%; group 3, 72.5%; and group 4, 29.9%. These differences also did not reach statistical significance (Fig. 1). In the 80 patients who had a late seizure, 26 seizures occurred in the context of medication reduction or stress, and in only five of these 26 cases did seizures persist. At the most recent follow-up, only 37 (13%) patients were not seizure free. Percentage recurrence rates for each subgroup are presented in Table 3.
Figure 1. Risk of having at least one seizure after 1 year of being seizure free, depending on the preoperative magnetic resonance imaging results and the location of the seizure onsets.
Download figure to PowerPoint
Table 3. Recurrence rate for each subgroup
|Center|| ||A||B|| |
| ||28.7||24.3|| |
|GTC|| ||Yes||No|| |
| ||33.3||22.9|| |
|Age at seizure onset|| ||<6 yr||6–15 yr||>15 yr|
|Years with seizures|| ||<15 yr||15–24 yr||<24 yr|
|Age at surgery|| ||<26 yr||27–36 yr||>36 yr|
|Auras|| ||Yes||No|| |
| ||26.1||30.9|| |
|Side|| ||Left||Right|| |
| ||25.5||28.5|| |
Univariate analysis revealed two factors that were predictive of seizure recurrence. Patients with preoperative GTC seizures were more likely to have a recurrence of their seizures after 1 year of being seizure free (p = 0.03; Fig. 2). On subgroup analysis, this was significant only for patients with neocortical onsets (p = 0.04), and not for patients with MTLE (p = 0.14). Age at surgery was also significant according to both the Cox model (p = 0.049) and a Kaplan–Meier log rank stratified by terciles (p = 0.004; Fig. 3). Patients operated on before age 26 years had a lower incidence of late recurrence. On subgroup analysis, this was significant only for MTLE (p = 0.01) but trended toward significance for patients with neocortical onsets (p = 0.08). Although age at seizure onset was not significant for the whole group, for patients with MTLE, later age at onset was significantly associated with seizure recurrence after 1 year for MTLE only by using the Kaplan–Meier log-rank by terciles but not the Cox model (p = 0.04).
Figure 2. Risk of having at least one postoperative seizure after 1 year of being seizure free, depending on the presence or absence of preoperative generalized tonic–clonic seizures.
Download figure to PowerPoint
Figure 3. Risk of having at least one postoperative seizure after 1 year of being seizure free, depending on the age at surgery by tercile.
Download figure to PowerPoint
We then performed a multivariate analysis using the presence of preoperative GTC seizures and age at surgery as variables. Each was an independent predictor of at least one recurrent seizure after 1 year of seizure freedom. This finding was significant both for the Cox model [presence of GTC (p = 0.004), age at surgery (p = 0.049)] and the Kaplan–Meier log rank by terciles [presence of GTC (p = 0.004), age at surgery (p = 0.004)]. However, age at surgery was significant only for patients with MTLE (p = 0.02).
- Top of page
In this large retrospective bi-institutional study, we found that as many as 28% of patients who are seizure free for 1 year after surgery will have a subsequent seizure. However, this finding does not mean that surgery failed, because a single late postoperative seizure did not necessarily predict further seizures. At last follow-up, only 13% of the patients were not Engel class I. Thus in counseling our patients, we advise that being seizure free at 1 year does not suggest that they will never have another seizure, but the chances of recurrent seizures developing is relatively low. We attribute this outcome stability to a combination of restarting previously discontinued medications, transience of seizure precipitants (stress, lack of sleep, fever), and the running-down phenomenon. Although the literature stresses the curative potential of epilepsy surgery, in at least half of patients who are “seizure free,” the maintenance of their medication regimen remains essential. It is more accurate to think of their seizures as “controlled” rather than “cured” with surgery. In spite of the potential for recurrent seizures, all patients who are seizure free, regardless of whether they are taking medications, will gain the benefits of decreased morbidity and mortality rates, and increased quality of life and employment rates (1,31,32).
Although some studies have suggested that outcome at 1 year is a fairly good predictor of long-term outcome (2,6,11,12,15), other studies have highlighted the potential for a late recurrence (14,18,24,27,28). Most studies tend to agree that late recurrences do not often become intractable (9,11,27) and are partially balanced by patients whose epilepsy runs down over time (30,33,34). Ultimately, seizure control does decrease with increasing follow-up. Engel et al. (13) reported that only 79% of patients who are seizure free for 1 year will remain seizure free, and only 63% will still be seizure free at 10 years. Yoon et al. (27) recently published a series demonstrating that the probability of late recurrence for those seizure free at 1 year was 28% at 5 years and 44% at 10 years. McIntosh et al. (28) found a similar dramatic decline in seizure control after surgery for MTLE. Of those patients who were seizure free for 1 year, only 47% remained seizure free at 5 years, and 41% remained seizure free at 10 years. Although our results confirm the ever-present risk of late recurrence, the rate of recidivism that we found is not quite as high as that in previous reports. One explanation may be the use of MRI scanning in all patients included in our series. Because prior studies included patients operated on before MRI scans were available, subtle lesions may have been missed (13,27,28). Although MRI results did not correlate with late recurrence in our study, resections of imaging abnormalities were likely more extensive in our study, with a lower chance of residual structural pathology that might have caused a late recurrence in other studies performed without the benefit of MRI scans on all patients. Our results mirror a recent meta-analysis of long-term outcome after epilepsy surgery that found persistently high rates of seizure control, particularly with temporal lobe surgery (4).
The most significant predictors of late recurrence after resective surgery by multivariate statistical analysis were the presence of GTC seizures in patients with neocortical epilepsy and later age of surgery in patients with MTLE. The presence of preoperative GTC seizures has previously been shown to predict recurrence after epilepsy surgery (3,24), even for patients who have been seizure free for a year (28), although contrary data exist (27). We found this observation to be true only of neocortical epilepsy and not of MTLE, but McIntosh et al. (28) reported that GTC seizures predict late recurrence after surgery for MTLE as well (28). Presumably, GTC seizures are a marker for more widespread cortical involvement and a distributed epileptogenic zone (35). By contrast, an intermediate-sized epileptogenic zone is thought to predict the running-down phenomenon, because a small amount of residual epileptogenic tissue may be capable of generating seizures only for a limited time (30,33). It is not clear why GTC seizures were a negative predictor only for neocortical epilepsy; the frequency of GTCs in this group was much higher than that in the MTLE group, potentially introducing bias.
A later age at surgery also has been described as a negative predictor of outcome in several studies of epilepsy surgery (2,5,7,10,36,37). Nevertheless, older patients do quite well after surgery and should not be eliminated from consideration for surgery merely because of their age (38,39). McIntosh et al. (28) also found later age at surgery to predict late recurrence after initially successful MTLE surgery on univariate analysis. Yoon et al (27), however, showed that the duration of preoperative seizures correlated with late recurrence. It is likely that older age at surgery and duration of preoperative epilepsy are related. Both were found to be predictors of late recurrence in a recent study by Jansky et al. (24). Their conclusion, supported by our data, is that early surgical intervention is important to optimize outcome. The average duration of epilepsy before surgery in our study was 22 years, which indicates that physicians are still reluctant to refer patients for epilepsy surgery early in their disease progression in spite of evidence of the benefits of early surgery.
Although we did not find any statistically significant difference in the rate of late recurrence in patients with normal MRI scans or neocortical epilepsy, the trend was in this direction. However, we emphasize that more than half of the patients with normal MRI and TLE had MTS on postoperative histology, and one third of the TLE patients with atrophy had normal histology. Hence, preoperative imaging may not be as good a predictor of late seizures as is postresection histology. The impact of preoperative imaging has not previously been examined specifically in patients who are initially seizure free, but good data exist on the impact of histology. McIntosh et al. (28) and Yoon et al. (27) both found that seizure recurrences after initially successful surgery were more common in patients with normal histology. Contrary data show a higher rate of late recurrence in patients with MTS or dysplasia compared with patients with normal histology among patients with MTLE (11,14,17,40), but this finding is not borne out in most other studies (12,16,33,37,41). In reports that did examine preoperative imaging as a predictor of outcome, the consensus is that patients with normal scans have a higher risk of recurrent seizures after surgery (6,11,33,42–44). However, these studies did not specifically examine patients who were initially seizure free and later had a seizure recurrence. We presume that patients with normal MRI scans are likely have more widespread epileptogenic zones that are more difficult to localize and resect completely, leading to a greater possibility of residual epileptogenic tissue that can elicit seizures triggered by medication withdrawal or stressful precipitating events (34,45). Another potential explanation for the lack of statistical significance in our study, in addition to the variable histology in patients with TLE and normal MRI scans, as mentioned earlier, may be the frequent use of subdural electrodes in patients with both normal and abnormal MRI scans. We would have thought that patients in whom subdural electrodes were used would also have a higher rate of late recurrence, because subdural electrodes are generally used to define a seizure focus in the absence of an imaging abnormality. However, our data did not support this assertion. Possibly the use of subdural electrodes permitted the identification of epileptogenic foci and functional cortex in patients with normal MRI scans, thereby increasing the extent of resection of the focus. Thus a normal MRI scan is not, in and of itself, a bad prognostic indicator if subdural electrodes can identify a regionalized focus for resection or if histology is abnormal. Patients in the subgroup that is seizure free for a year likely have such a focus. Another possibility is the modest sample size of certain subgroups, particularly patients with neocortical onsets and normal MRI scans.
Our finding that patients with neocortical epilepsy do not have a statistically significant increase in late relapse is not altogether surprising. Although patients with neocortical surgery generally have a higher rate of failure compared with MTLE patients (4,9,13,21–23,26), most of these failures likely occur within the first year and would be excluded from our study Other reports, however, do not find such a difference (9,17,20). Nevertheless, in a smaller single-institution study with a design similar to ours, Yoon et al. (27) also found that patients with extratemporal resections do not have a higher rate of recidivism if they have been seizure free for 1 year. Although we found a slightly higher rate of recurrent seizures in the extratemporal group, our modest sample size clearly limited the power of this trend.
Antiepileptic medication reduction was associated with seizure recurrence in one third of our patients in whom seizures recurred. However, seizures persisted in only a small fraction of these cases. In previous studies, medication reduction has been thought to play a more important role in seizure recurrence and the reduction in seizure-free rates after surgery (27,29,46,47). In a study that specifically examined patients who were initially seizure free after surgery, Yoon et al. (27) found that as many as 51% of seizure recurrences were precipitated by medication withdrawal. However, in another such study, McIntosh et al. (28) did not report an association between medication withdrawal and late recurrence. This inconsistency may depend on the involvement of the epileptologist in the initiation and rate of the taper, which may vary between studies.
In summary, we found that as many as one third of patients who are seizure free for 1 year after epilepsy surgery may eventually have at least one seizure. However, only 13% will redevelop intractable epilepsy. The etiology of late recurrence is multifactorial and in most cases represents a transient dysequilibrium in seizure control after surgery. The finding that preoperative GTC seizures and late age at surgery are preoperative predictors of late recurrence indicates the importance of patient selection and early surgery for persistent seizure control.