SEARCH

SEARCH BY CITATION

Keywords:

  • Temporal lobe epilepsy;
  • Epilepsy surgery;
  • Seizure recurrence;
  • Seizure remission

Summary

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References

Purpose: Temporal lobectomy is a well-established treatment for refractory temporal lobe epilepsy, yet many patients experience at least one seizure postoperatively. Little is known about the prognostic significance of the time from surgery to first seizure relapse in predicting long-term outcome.

Methods: In a retrospective analysis of patients who reported at least one complex partial seizure (CPS) or generalized tonic–clonic seizure (GTCS) after anterior temporal lobectomy (n = 268), we used a nominal response logistic model to predict the odds ratio (OR) of a seizure outcome based on length of the latency period from surgery to first postoperative seizure. A modified Engel outcome class scheme was used. We controlled for factors known to influence postoperative outcome, including history of tonic–clonic seizures, intelligence quotient (IQ), preoperative seizure frequency, magnetic resonance imaging (MRI) findings, and history of febrile convulsions.

Results: In the univariate analysis, the latency from surgery to the first postoperative disabling seizure was significantly associated with long-term outcome. Longer latency was associated with higher odds of being seizure-free or improved (modified Engel’s classes 1, 2, and 3) relative to the unimproved state (class 4) (p < 0.001, 0.001 and 0.004, respectively). Conversely, a shorter latency increased the likelihood of achieving the worst prognosis (class 4) relative to class 1 (p < 0.001). Multivariate analysis yielded similar results.

Discussion: Latency to the first postoperative seizure predicts long-term outcome, with short latencies portending poor prognosis and long latencies portending a good prognosis. This information can be used for patient counseling and may influence decisions regarding reoperation.


Background

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References

Temporal lobectomy is a well-established treatment for medically refractory temporal lobe epilepsy (TLE) and carries a favorable prognosis, with 60–70% of patients achieving terminal remission (no seizures in last year of follow-up) following surgery (McIntosh et al., 2001). However, a substantial portion of patients will experience at least one seizure postoperatively. A literature review by McIntosh et al. (2001) found seizure freedom rates ranging from 53–80% at one postoperative year and 52–58% at two postoperative years. Even as late as 10 years after surgery, de novo recurrences are reported, with approximately 60% of patients having at least one recurrence in the first 10 years (Sperling et al., 2008). Given that seizure relapse occurs in many patients after surgery, the clinical significance of such a recurrence can be questioned. Whether a single seizure represents failure of surgery, implies that further seizures may occur, or even raises the consideration of reoperation is a critical concern that warrants further investigation.

Previous studies have analyzed the relationship between the occurrence of “neighborhood” seizures (usually seizures within 1–4 weeks after surgery) and long-term outcome, concluding that postoperative seizures within 1 month of surgery are associated with future recurrences (Malla et al., 1998; Radhakrishnan et al., 2003; McIntosh et al., 2005). However, no study of temporal lobectomy patients has examined whether the length of time to the first postoperative seizure can be used to predict outcome. Several authors who used the length of time to first postoperative seizure, survival analysis, as a measure of surgical outcome did not explore its possible power to predict long-term prognosis (McIntosh et al., 2004; Keleman et al., 2006).

In this article, we report an analysis of the relationship between the number of days to first postoperative seizure and long-term outcome, controlling for major prognostic factors. We hypothesize that a shorter latency period will correlate with a poor seizure outcome, and a longer time to first seizure will be predictive of a good long-term outcome.

Methods

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References

Subjects

Data were drawn from a surgical database, prospectively maintained since 1986. The cohort for analysis included patients who failed to respond to a minimum of two antiepileptic medications and subsequently had an anterior temporal lobectomy. Surgery was performed by three different surgeons over the years. A medical history, physical examination, neurologic examination, interictal and ictal electroencephalography (EEG), magnetic resonance imaging (MRI), positron emission tomography (PET), and neuropsychological testing were all performed on each patient. Our evaluation protocol has evolved over the years and has been published previously (Chandrasekar et al., 2008).

We then applied the following inclusion criteria for the analysis: (1) no additional brain surgery for epilepsy; (2) minimum 1-year follow-up since surgery; (3) at least one disabling seizure with loss of awareness since surgery (complex partial or tonic–clonic); (4) no pseudoseizures postoperatively; and (5) preoperative history of complex partial seizure (CPS), not exclusively simple partial seizures prior to surgery. Simple partial seizures, defined as auras without motor phenomena or loss of consciousness, were not considered disabling regardless of whether they occurred preoperatively or postoperatively and, thus, were not included in the outcome analyses. Early perioperative simple partial motor seizures occurred in only one patient, and these were not considered disabling. We excluded pseudoseizures because of underlying ambiguity regarding the patient’s neurologic condition. The Thomas Jefferson University Institutional Review Board approved this project.

Clinical variables

Clinical data included age at surgery, gender, full-scale intelligence quotient (IQ) at preoperative assessment, preoperative and postoperative seizure history, preoperative MRI findings, history of tonic–clonic seizures, history of febrile convulsions, date of surgery, date of first recurrence, duration of follow-up since time of surgery, and current seizure outcome. There are four current seizure outcome classes reflecting condition in the last year of follow-up, modified from Engel’s scheme and assessed at time of last visit: Class 1 is no disabling seizures, defined as complex partial seizures or tonic–clonic seizures, which qualifies as terminal remission; class 2 is disabling seizures on <3 days per year or exclusively nocturnal seizures; class 3 is ≥80% reduction in disabling seizures; and class 4 is <80% reduction in disabling seizures. Postoperative assessments were conducted at office visits and during standardized telephone interviews to evaluate seizure recurrence. Latency to first postoperative disabling seizure was calculated for each patient, subtracting date of surgery from date of first seizure recurrence after surgery. Several clinical factors have previously been shown to correlate with postoperative prognosis. A history of tonic–clonic seizures (Berg et al., 1998; Janszky et al., 2005a), low IQ (Chelune et al., 1998), and high preoperative seizure frequency (Foldvary et al., 2000) are associated with a poor outcome, whereas structural lesion seen on MRI (including mesial temporal sclerosis, MTS) (Berkovic et al., 1995; Radhakrishnan et al., 1998) and a history of febrile convulsions (Salanova et al., 1994; Janszky et al., 2003) positively correlate with a favorable long-term prognosis. These variables were included in the analysis. We categorized MRI findings into normal, mesial temporal sclerosis, or extratemporal radiologic abnormality outside of the temporal lobe (e.g., gliosis, heterotopia, and so on). Whether medications influenced relapse could not be determined, given that >90% of subjects were taking an antiepileptic drug (AED) at the time of first relapse. Clinical practice varied sufficiently with regard to alteration in AED doses such that no analysis is possible.

Data analysis

The four possible seizure outcome classes were analyzed in a nominal response logistic model (also known as generalized logit model) with either the worst outcome, class 4, or the best outcome, class 1, as the reference category. This model is the generalization of the standard logistic regression model to the responses with more than two categories. In this model, the log generalized odds (generalized logits) are modeled as the linear function of covariates. The generalized log odds or logits are the log transformed ratios of the probability of nonreference category divided by the probability of the reference category. Therefore, they have the usual interpretation of odds being in the nonreference category relative to the reference category.

The univariate model was used to evaluate the association between the four-category outcome and the latency to first seizure. The multivariate model also controlled for the following clinical variables: MRI findings, history of febrile convulsions, history of preoperative tonic–clonic seizures, full-scale IQ, and preoperative seizure frequency. Confidence intervals (CIs) were calculated using the Wald Statistic. Statistical analyses were performed in SAS 9.1 (SAS Institute Inc., Cary, NC, U.S.A.).

Results

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References

Two hundred sixty-eight patients were eligible for the present analysis of a total of 677 patients who had a temporal lobectomy. Three hundred seven patients have not experienced any disabling seizures (complex partial or tonic–clonic) since surgery and were ineligible for inclusion. Fifty-four patients were excluded because they had additional brain surgery, two patients were excluded because they had only simple partial seizures prior to surgery, 17 were excluded due to follow-up duration <1 year, and 11 were excluded because they had postoperative pseudoseizures. Eighteen patients had incomplete data and were also excluded. The mean age at surgery was 36.2 ± 11 years. Thirty percent of patients with postoperative seizures had only a single recurrence; the remaining 70% had two or more postoperative seizures. Characteristics for each outcome group are summarized in Table 1.

Table 1.   Patient characteristics
CharacteristicClass 1 (n = 134)Class 2 (n = 55)Class 3 (n = 59)Class 4 (n = 25)
  1. MTS, mesial temporal sclerosis; TC, tonic–clonic seizures.

Median days to first seizure: (range)218 (2, 5015)391 (1, 5915)107 (2, 4941)46 (2, 391)
Median preop full-scale IQ: (range)89 (56, 135)92 (68, 126)89 (70, 124)83 (66, 106)
Median preop seizure frequency: (per month) (range)5 (0.2, 255)5 (0.2, 75)6 (0.2, 181)6 (0.3, 120)
MRI findings: % normal/% MTS/% extratemporal abnormality  20%/57%/23%22%/62%/16%24%/37%/39%  32%/32%/36%
History of febrile convulsions: N (%)53 (40%)13 (26%)12 (20%)5 (20%)
History of preop TC seizures: N (%)50 (37%)17 (34%)27 (46%)11 (44%)

Prediction of outcome

In the univariate analysis, the latency from surgery to the first postoperative disabling seizure was significantly associated with the generalized odds of class 1, 2, or 3 outcomes as opposed to class 4. Table 2 shows the odds ratios (ORs) for class 1, 2, or 3 versus class 4 and class 2, 3, or 4 versus class 1 associated with a 10-fold increase in the latency from surgery to the first postoperative disabling seizure. A significantly longer latency was associated with higher odds of class 1, 2, or 3 relative to 4 (p < 0.001, <0.001, and 0.004, respectively). The odds of class 2 or 3 versus class 1 outcome were not significantly associated with the latency from surgery to the first postoperative disabling seizure.

Table 2.   Outcome versus latency to first postoperative seizure: univariate analysis (n = 268)
 Outcome class comparisonOdds ratio(95% CI)p-value
  1. aThe number of days to the first seizure was log-transformed for this analysis.

Days to first seizurea1 vs. 43.05(1.75, 5.31)<0.001
2 vs. 44.01(2.13, 7.55)<0.001
3 vs. 42.40(1.33, 4.33)0.004
Days to first seizurea2 vs. 11.32(0.87, 1.98)0.19
3 vs. 10.79(0.54, 1.14)0.21
4 vs. 10.33(0.19, 0.57)<0.001

Figure 1 shows this effect expressed in simple time intervals. A longer latency to the first postoperative seizure is associated with greater chance of terminal remission and lower probability of having a second seizure. For patients experiencing their first postoperative seizure in the first 3 months, 4–6 months, 7–12 months, 1–3 years, and after 3 years, the proportions having a second seizure were 82% (95% CI 0.0751), 77% (95% CI 0.1371), 74% (95% CI 0.1368), 69% (95% CI 0.1315), and 37% (95% CI 0.1307), respectively. The chance of being in terminal remission when the study was censored was, respectively, 43% (95% CI 0.0947), 60% (95% CI 0.1547), 53% (95% CI 0.1513), 47% (95% CI 0.14), and 61% (95% CI 0.1319). There was no difference in chance of achieving terminal remission for patients who relapsed in the first month compared with those who relapsed in months 2 and 3 after surgery (p = 0.4168, chi-square = 0.659). Sixteen (37.2%) of 43 patients who initially relapsed during months 2 and 3 achieved terminal remission, and 27 (47.4%) of 57 patients who experienced their first postoperative seizure during month 1 achieved terminal remission. Therefore, these data are grouped.

image

Figure 1.   Seizure outcomes based on timing of initial relapse. Red, percentage who experienced a second seizure; blue, percentage who achieved terminal remission.

Download figure to PowerPoint

For latency to the first postoperative seizure, the results from the multivariate generalized logit model, adjusted for other covariates, were similar to those obtained in the univariate analysis. In addition to the latency to the first postoperative seizure, only the MRI findings were an overall significant predictor of outcome (p = 0.014). History of febrile convulsions approached statistical significance (p = 0.073). Table 3 shows the multivariate analysis results for the model examining the odds of better than class 4 outcomes, and Fig. 2 displays these data. In particular, a 10-fold increase in latency to the first postoperative seizure implied a 3.5-fold increase in generalized odds of class 1 versus class 4, a 4.9-fold increase in generalized odds of class 2 as opposed to class 4, and a 2.7-fold increase in generalized odds of class 3 compared to class 4. A second multivariate analysis was performed examining the generalized OR for class 1 outcome (terminal remission). Only the odds of class 4 versus class 1 outcome were significant, specifically only for latency to first seizure (p < 0.001) and presence of MTS on MRI scan (p = 0.009). No other clinical variables predicted terminal remission.

Table 3.   Multivariate analysis assessing odds ratios for class 4 outcome
Clinical variableAnalysis parameterOutcome class comparisonOdds ratio(95% CI)p-value
  1. TC, tonic–clonic seizure; MTS, mesial temporal sclerosis.

  2. aThe number of days to first seizure were log-transformed for this analysis.

  3. bThe preop seizure frequency was log-transformed for this analysis.

Days to first seizurea 1 vs. 43.48(1.89, 6.39)<0.001
2 vs. 44.93(2.45, 9.92)<0.001
3 vs. 42.68(1.42, 5.05)0.002
MRI findingsNormal versus MTS1 vs. 40.19(0.05, 0.66)0.009
2 vs. 40.14(0.04, 0.57)0.006
3 vs. 40.33(0.09, 1.24)0.099
Abnormal (extratemporal) versus MTS1 vs. 40.26(0.08, 0.84)0.024
2 vs. 40.13(0.03, 0.53)0.004
3 vs. 40.59(0.17, 1.99)0.39
History of febrile convulsionsNo versus Yes1 vs. 40.46(0.15, 1.46)0.19
2 vs. 40.94(0.25, 3.47)0.92
3 vs. 41.02(0.29, 3.58)0.97
History of preoperative TC seizuresNo versus Yes1 vs. 40.81(0.31, 2.13)0.67
2 vs. 40.81(0.27, 2.42)0.70
3 vs. 40.63(0.23, 1.74)0.37
Preoperative seizure frequencyb 1 vs. 41.16(0.46, 2.91)0.75
2 vs. 41.26(0.45, 3.52)0.66
3 vs. 41.46(0.56, 3.85)0.44
Preoperative IQ<80 versus ≥801 vs. 40.48(0.17, 1.34)0.16
2 vs. 40.23(0.06, 0.83)0.025
3 vs. 40.47(0.16, 1.40)0.18
image

Figure 2.   Predicted generalized log odds of class 1 versus class 4. Red, mesial temporal sclerosis on MRI; blue, extratemporal abnormality on MRI; green, normal MRI.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References

The major finding of this study is that latency to the first seizure recurrence is an independent and important predictor of long-term outcome after temporal lobectomy. A shorter latency period between epilepsy surgery and the first postoperative seizure correlates with higher odds of experiencing continued seizures and a worst possible outcome class. The longer a patient remains seizure free following surgery, the greater the likelihood of being in terminal remission (no seizures in preceding year) and the better the chance of at least an 80% reduction in seizure frequency. In particular, our data showed that an initial relapse beyond three postoperative years is associated with a very low probability of class 4 outcome, that is, <80% seizure reduction.

Recurrence suggests the presence of residual epileptogenic cortex, with a preserved tendency to have continued seizures. Germano et al. (1994) studied anteromedial temporal lobectomy patients who underwent reoperation for continued seizures. MRI performed prior to the second procedure revealed residual mesial temporal structures in all patients. Following the reoperation and removal of the remaining mesiotemporal tissue, those patients who did not become seizure-free had EEG abnormalities that expanded to other areas of the brain, suggesting that the epileptogenic zone in this population was not confined to the temporal lobe. Two subsequent studies confirm these results. First, in a prospective, randomized clinical trial, Wyler et al. (1995) demonstrated that anterior temporal lobectomy with total hippocampectomy, compared with partial hippocampectomy, resulted in much higher rates of seizure freedom. Secondly, Hennessy et al. (2000) found that among patients who remained refractory after temporal lobectomy, most exhibited extrahippocampal and extratemporal epileptogenesis. The common denominator among these studies is that a sufficient amount of epileptogenic tissue that remains after surgery will generate continued seizures, regardless of location within the brain. Early recurrence indicates that the epileptogenic network was not disrupted by surgery, perhaps due to inadequate resection of epileptogenic cortex.

What can be surmised regarding patients with a long latency until the first postoperative seizure? Clearly, epileptogenic tissue remained following surgery, but either a sufficient amount was removed or a critical network was interrupted so that seizures could no longer be generated. That seizures ultimately recurred argues for the possibility of a progressive process, whereby epileptogenicity was reestablished, resulting in a more benign form of epilepsy with a tendency toward infrequent seizures. McIntosh and Berkovic (2006) suggest that these patients may have an underlying, diffuse network of neurons with low-grade epileptogenic potential, which has been unmasked by surgery. Alternatively, perhaps the entire epileptogenic zone was resected and late seizure recurrences represent gradual maturation of a second, more benign seizure focus (Eliashiv et al., 1997; McIntosh & Berkovic, 2006). It is possible that a critical number of pathologic neurons is required for generation of frequent, disabling seizures and, following ATL, that the critical number of neurons is no longer present. Lastly, it is possible that surgery caused a new epileptogenic lesion, which undergoes a prolonged maturational process. In the present study, patients who went several years without seizures had a small chance of reverting to class 4, that is, developing frequent seizures. Therefore, it would appear that the probability of retained pathologic cortex sufficient to generate severe epilepsy is low in class 1 patients.

Length of time to the first postoperative seizure as a predictor of surgical outcome following temporal lobectomy has never previously been studied in this way, to our knowledge. Yoon et al. (2003) studied timing of seizure relapse and outcome in patients who underwent any type of brain surgery for epilepsy, not exclusively temporal lobectomy. In contrast to the present study, the population included patients who initially remained seizure free for the first postoperative year, may or may not have undergone reoperation, and were at least 3 years of age. Of the patients who relapsed beyond one postoperative year (n = 63), the time to first recurrence did not predict whether a patient would have a mild versus moderate recurrence, defined as one or less, or more than one, seizure per year. They did find, however, that a longer seizure-free interval was associated with reduced likelihood of having a severe recurrence, similar to the present results.

Why some patients enter long-term remission after experiencing postoperative seizures is not understood. Epidemiologic studies of epilepsy reveal spontaneous seizure remission as time passes (Beghi & Sander, 2008). The running-down phenomenon (Salanova et al., 1996), in which remission takes some time, has been observed after surgery, and some patients with refractory seizures enter spontaneous remission with prolonged medical treatment (Callaghan et al., 2007). This phenomenon may contribute in part to favorable responses after surgery.

A few studies looked at specific intervals, such as 1 month, 1 year, or 2 years, in relation to outcome and found that those patients who relapsed within the first year tended to have a worse outcome (Radhakrishnan et al., 2003; McIntosh et al., 2005; Keleman et al., 2006). Salanova et al. (1999) also evaluated seizure freedom rates at yearly intervals and demonstrated that those who remain seizure-free within the first postoperative year were less likely to have reported seizure recurrence at the time of last-follow-up. In the current study, however, we sought to further delineate the prognostic significance of a seizure relapse by excluding seizure-free patients and not arbitrarily choosing a time period for analysis. The multivariate analysis and calculation of ORs is also novel.

The population we chose for analysis includes only those patients who experienced at least one disabling seizure (CPS or tonic–clonic) following surgery, since the purpose of this analysis was to examine the relationship between latency to first seizure and outcome. There are important advantages to excluding seizure-free patients: (1) those who remain seizure-free postoperatively generally have milder forms of epilepsy (Jeong et al., 2005), which may reflect a different underlying pathophysiology; and (2) our statistical analysis is perhaps based on a more homogeneous population, which reduces the chance of outliers skewing the data.

We further validated our results by controlling for prognostic factors that have previously been established as predictive of surgical outcome. Only MRI was associated with terminal remission in patients who relapsed, in addition to seizure latency. The other variables, although predictive of outcome in general for patients who have ATL, did not appear related to outcome in patients who experienced relapse. This may be related to study design; this study provides a different perspective than the usual outcome study, by focusing attention on the subset of patients who have had at least one relapse rather than all patients who have had surgery.

This study has several limitations. Patients were followed for different lengths of time since surgery and, thus, terminal remission was assessed at different postoperative times. This effect was managed, however, by analyzing only the number of days from surgery to the first seizure recurrence and by excluding patients whose duration of follow-up was <1 year after the first postoperative seizure when assessing terminal remission. It might be desirable to study a group of patients with a homogeneous duration of follow-up; sample size limitation prevented this analysis. In addition, another analysis might consider a longer time period for terminal remission, but the 1-year time frame has been used widely in previous literature (Berg et al., 1998; McIntosh et al., 2001). Lastly, postoperative medications were managed clinically and variably; however, at least one study (Berg et al., 2006) suggests that medications have a limited role with regard to postoperative seizure recurrence.

Many reports emphasize the role of preoperative tests and clinical features in predicting outcome. Yet after surgery, many patients wonder what the future will bring, both in terms of seizure outcome and quality of life. It is important for patients and physicians to have confidence in the future, based on present information. Several studies already have shown that as seizure frequency decreases after surgery, or length of time in remission increases, patients report a significant increase in quality of life (Wass et al., 1996; Spencer et al., 2007; Téllez-Zenteno et al., 2007). Determination of seizure outcome based on objective clinical data, however, is less well-established (Spencer et al., 2005). Some patients with good preoperative prognostic features relapse and some with bad features do not. Early seizure recurrence suggests the presence of residual epileptogenic tissue and, thus, perhaps an indication for early reoperation. Conversely, the more time that passes until patients experience their first postoperative seizure, the better their long-term prognosis will be, and the more confidence they can have in their future. The data more robustly predict lack of poor outcome in patients with a long latency to first postoperative seizure. Therefore, physicians can confidently assure patients with a late recurrence an ultimate good prognosis.

The present study provides some evidence for preliminary recommendations with regard to clinical management. Prompt adjustment of medical therapy or reoperation should be considered soon for patients who relapse within a few months of surgery, since chance of a poor outcome is high. Those patients who have late seizure recurrence, particularly after 3 years, may be better served by a more conservative approach that delays consideration of changing medications or additional surgery. Additional research is needed to further refine outcome prediction after surgery. Although postoperative auras (Chandrasekar et al., 2008), postoperative AED regimen changes (Schiller et al., 2000; Janszky et al., 2005b), and postoperative EEG results (Tuuneinen et al., 1994; Patrick et al., 1995; Radhakrishnan et al., 1998; Hildebrandt et al., 2005) have been reviewed, other variables may also yield information of value.

Disclosure

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References

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. There are no relevant disclosures.

References

  1. Top of page
  2. Summary
  3. Background
  4. Methods
  5. Results
  6. Discussion
  7. Disclosure
  8. References