OUP user menu

Temporal lobe epilepsy with hippocampal sclerosis: predictors for long-term surgical outcome

J. Janszky, I. Janszky, R. Schulz, M. Hoppe, F. Behne, H. W. Pannek, A. Ebner
DOI: http://dx.doi.org/10.1093/brain/awh358 395-404 First published online: 5 January 2005

Summary

Temporal lobe epilepsy (TLE) accompanied by hippocampal sclerosis (HS) is the type of epilepsy most frequently operated on. The predictors for long-term seizure freedom after surgery of TLE-HS are unknown. In this study, we aimed to identify prognostic factors which predict the outcome 6 months and 2, 3 and 5 years after epilepsy surgery of TLE-HS. Our working hypothesis was that the prognostic value of potential predictors depended on the post-operative time interval for which the assessment was made. We included 171 patients (100 females and 71 males, aged 16–59 years) who had undergone presurgical evaluation, including video-EEG, who had had MRI-defined HS, and who had undergone temporal lobectomy. We found that secondarily generalized seizures (SGTCS) and ictal dystonia were associated with a worse 2-year outcome. Both these variables together with older age and longer epilepsy duration were also related to a worse 3-year outcome. Ictal limb dystonia, older age and longer epilepsy duration were associated with long-term surgical failure evaluated 5 years post-operatively. In order to determine the independent predictors of outcomes, we calculated multivariate analyses. The presence of SGTCS and ictal dystonia independently predicted the 2-year outcome. Longer epilepsy duration and ictal dystonia predicted the 3-year outcome. Longer epilepsy duration (P = 0.003) predicted a poor 5-year outcome. Conclusively, predictors for the long-term surgical results of TLE with HS are different from those variables that predict the short-term outcome. Epilepsy duration is the most important predictor for long-term surgical outcome. Our results strongly suggest that surgery for TLE-HS should be performed as early as possible.

  • epilepsy surgery outcome
  • prognosis
  • duration
  • generalized tonic-clonic seizures
  • ictal dystonia
  • AED = anti-epileptic drugs
  • CI = confidence interval
  • HS = hippocampal sclerosis
  • IED = interictal epileptiform discharges
  • OR = odds ratio
  • ROC = receiver operating characteristic
  • SGTCS = generalized tonic-clonic seizures
  • TLE = temporal lobe epilepsy

Introduction

Temporal lobe epilepsy (TLE) is the most common type of epilepsy requiring surgical treatment (Engel et al., 1997). The majority of patients with TLE have hippocampal sclerosis (HS) as a pathological abnormality underlying TLE (Babb and Brown, 1987). TLE accompanied with HS (TLE-HS) has been recently recognized as a well-defined epilepsy syndrome (Wieser et al., 2004). Surgical treatment of TLE (Wiebe et al., 2001) and especially TLE-HS (Radhakrishnan et al., 1998) has a favourable prognosis, 60–70% of patients become seizure-free after temporal lobe resections (McIntosh et al., 2001).

Determination of prognostic factors for TLE-HS surgery is an important influence when counselling our patients in everyday practice. In addition, the identification of prognostic factors may improve general understanding of the pathophysiology of surgical failure and of spatial extension of the epileptogenic zone. Many studies have looked at predictive factors for favourable outcome, but most included patients with a variety of temporal lobe pathologies and not patients with just TLE-HS. The most frequently reported predictors for favourable outcome of TLE surgery are unilateral temporal pathology on MRI (Berkovic et al., 1995; Radhakrishnan et al., 1998; McIntosh et al., 2001) and unilateral interictal epileptiform discharges (IED) (Bengzon et al., 1968; Blume et al., 1994; Holmes et al., 1997; Radhakrishnan et al., 1998; Schulz et al., 2000). Age at operation (Blume et al., 1994; Guldvog et al., 1994), epilepsy duration (Guldvog et al., 1994; Specht et al., 1997) or febrile seizures (Abou-Khalil et al., 1993; Blume et al., 1994; Janszky et al., 2003) are less frequently reported predictors.

The main goal of epilepsy surgery is to provide long-term seizure freedom. However, because seizure outcome has typically been assessed only after the first or second post-operative years, very little is known as to what occurs in the subsequent 5–20 years (McIntosh et al., 2001; Yoon et al., 2003). The few long-term outcome studies that exist have found that patients relapse after many years of seizure freedom (Berkovic et al., 1995; Foldvary et al., 2000); thus, the long-term outcome is worse than the short-term and 48–55% of patients did not become seizure-free >5 years after the operation (McIntosh et al., 2001).

The cause for long-term worsening of the surgical outcome is unclear. Previous outcome studies were not able to answer this question as there is only one study (Berkovic et al., 1995) which determines the prognostic factors separately for the >5-year outcome and this study exclusively focused on the MRI findings. That study found that the presence of MRI abnormality highly predicts the long-term outcome. Another study (Yoon et al., 2003) aimed at evaluating the likelihood of risk factors for seizure recurrence in patients initially seizure-free after epilepsy surgery (various epilepsy syndromes and aetiologies were included). They found that longer duration of epilepsy and normal histopathological examination predicted late relapse after a 1-year seizure-free period. However, the minimum follow-up criterion in that study was 3 years and the EEG and neuroimaging data were not evaluated.

Only three studies have examined the outcome of surgery of TLE-HS (Kilpatrick et al., 1999; Hennessy et al., 2000b; Hardy et al., 2003). All of them, however, defined HS histopathologically, and not according to the pre-operative neuroimaging findings. Prediction of the post-operative outcome, however, requires pre-operative data. Moreover, only two of the three studies met the criteria of the high standards of modern presurgical evaluations (Kilpatrick et al., 1999; Hardy et al., 2003) such as high-resolution MRI and video-EEG recordings for all patients included in the study (Rosenow and Lüders, 2001). In addition, none of these three studies tried to identify predictive factors for long-term seizure (≥2 years) freedom. Identification of prognostic factors for long-term outcome has great importance in TLE-HS. Spencer et al. (2003) found that patients with mesial TLE who were seizure-free 1-year after the operation had a 24% relapse rate compared with neocortical epilepsy where the relapse rate in initially seizure-free patients was only 4%. Hennessy et al. (2000a) also found that the relapse in patients having initial 1-year seizure-free periods post-operatively occurred significantly more in TLE-HS than in other epilepsy syndromes.

In the present study, we aimed to identify prognostic factors which predict the outcome 6 months and 2, 3 and 5 years after epilepsy surgery of TLE with MRI-defined HS. Our working hypothesis was that long-term and short-term outcome may be influenced by different factors, assuming that the prognostic value of potential predictors depends on the post-operative time interval for which the assessment is made (Engel et al., 1993; Blume et al., 1994).

Methods

Presurgical evaluation at the Epilepsy Surgery Department of the Epilepsy Centre Bethel

Since observed variables were based on presurgical evaluation of epileptic patients, our procedure is briefly described below.

A detailed clinical history was taken from patients who were considered possible candidates for epilepsy surgery. Therapy resistance to first-line anti-epileptic drugs (AED) was evaluated. As a rule, high-resolution MRI was performed. With most patients, MRI pictures were made using a Siemens Magnetom Impact 1.5 T scanner, and included T1-weighted three-dimensional volume, protondensity, FLAIR and T2-weighted images.

Patients underwent continuous video-EEG monitoring lasting 2–10 days. In all patients, psychiatric and neuropsychological examination, as well as a social assessment, was performed. Findings from presurgical evaluation were discussed at a multidisciplinary case conference, where decisions were made concerning the possibility and type of surgery.

Patients who underwent epilepsy surgery were re-examined 6 months and 2 years later with an assessment of the seizure outcome as well as with an evaluation of the psychiatric, neuropsychological and social status. Three and 5 years after the operation, a specific questionnaire was sent to patients evaluating the seizure situation.

Our protocol is for patients to receive AED for a minimum of 2 years post-operatively. The dosage and type of AED remain unchanged unless the patients report side-effects. Two years after successful surgery, the option of a slow AED reduction and finally complete AED discontinuation is tailored to the individual patient.

Patient selection

For this study, we included all patients who underwent presurgical evaluation at our centre, aged >16 years, who had complex partial (psychomotor) seizures revealed by ictal video-EEG recordings, who were shown to have had HS on the MRI, and who had had a temporal lobectomy between 1993 and 2002 and who returned for at least the 6-month follow-up examination. Hippocampal sclerosis was defined if both hippocampal atrophy and T2 signal abnormalities were present on visual inspection of MRI pictures. Patients with bilateral HS or dual pathology (HS accompanied by other epileptogenic lesion) were excluded. A total of 171 patients met inclusion criteria. All the patients had a detailed clinical history, MRI and long-term video-EEG with ictal and interictal recordings.

Data collection

The following variables were investigated: (i) age at operation; (ii) sex; (iii) age at epilepsy onset; (iv) duration of epilepsy; (v) history of febrile seizures; (vi) seizure frequency; (vii) presence of secondarily generalized tonic-clonic seizures (SGTCS); (viii) presence of frequent SGTCS (occurring at least once a month); (ix) history of status epilepticus; (x) ictal limb dystonia; (xi) postictal aphasia; (xii) unilateral IED; (xiii) presence of contralateral seizure pattern; and (xiv) ‘standard’ temporal lobectomy.

The first nine clinical variables were ascertained by asking the patients and, in most cases, their relatives at admission to the presurgical unit. This history was taken by physicians blinded to goals of this study. Patients' previous medical charts were also reviewed. If the patient had had a non-febrile status epilepticus, a history of status epilepticus was defined. A history of febrile status epilepticus was not handled as a separate variable; this was included into the term ‘history of febrile seizures’. Data on EEG and the seizure semiology were determined by non-invasive continuous video-EEG monitoring. Ictal limb dystonia was considered present if sustained forced unnatural posturing of one upper extremity on one side of the body with a rotation component in the arm was present during the seizure (Kotagal et al., 1989). Postictal aphasia was defined according to previous studies (Gabr et al., 1989). In the present study, we did not distinguish whether these well-known lateralizing signs lateralized the epileptogenic region correctly or not. A 32–64 channel EEG recording was used. Electrodes were placed according to 10–10 systems; the number of electrodes and their placement varied individually taking consideration of the suspected epileptogenic region and side. The location and frequency of IED were assessed by visual analysis of interictal EEG samples of 2-minute duration every hour. Ictal EEG recordings were stored in separate files and were evaluated independently from the interictal data. Unilateral IED were defined if at least 95% of interictal discharges appeared over one temporal lobe.

‘Standard anterior temporal’ lobectomy was carried out up to a tempolateral distance of ≥3 cm from the temporal pole with maximal resection of the mesiotemporal structures. The keyhole resection combines the maximum resection of the mesiotemporal structures with an anteriotemporal resection of 1–2.5 cm from the temporal pole.

Selection of variables

The variables mentioned above were chosen according to their presumed importance in presurgical evaluation or if they were previously reported as a predictive factor for TLE surgery. The results of psychiatric, neuropsychological, and sociological assessments were not included.

Outcome assessment

Outcome was assessed at regular post-operative examinations (0.5 and 2 years after the operation), while the long-term outcome was evaluated by questionnaires sent 3 and 5 years after the operation. We evaluated only seizure outcome; psychiatric, social and neuropsychological outcomes were ignored. For this study, we divided the patients into two categories: (i) seizure-free outcome or (ii) non-seizure-free outcome. Patients with non-disabling auras without other seizures were considered to be seizure-free. Seizure freedom was defined if the patient was seizure-free from the operation to the last outcome assessment, or was seizure free for ≥2 years at the time of outcome assessment.

Statistical methods

Univariate logistic regression analyses were used to assess the prognostic importance of the clinical variables. Odds ratios (OR) were calculated for being seizure-free at 6 months and at 2, 3 and 5 years after the operation. To test the difference in prediction for a certain variable between the four different time points, we used a generalized linear model with repeated logistic design, Proc Genmod in SAS 8.02 (SAS Institute, Inc., Cary, NC, USA), which allowed us to investigate differences in a binary outcome given that the same subjects are investigated at different time points. We also performed a forward selection on the variables at each time point in order to determine the independent predictive factors for a favourable outcome. Variables with P < 0.05 were entered and variables with P > 0.10 were removed from the model. To describe the predictive ability of the combination of the variables proven to be independent predictors in this study, we constructed receiver operating characteristic (ROC) curves, and the area under the curve was calculated using STATA 8.0 (StataCorp, Lakeway Drive, College Station, TX, USA).

Results

Our inclusion criteria were met by 171 patients. The mean age of these patients was 33.1 ± 10 years (range 16–59 years). The age at epilepsy onset ranged from 1 to 37 years. There were 100 females and 71 males. A total of 104 patients (61%) underwent a standard lobectomy, while the remaining patients had keyhole resections. In 96 patients, the resection was on the left (56%) and, in the other 75 patients, it was on the right side. Histological results were available in 166 of the 171 patients. HS was confirmed in all patients. There were 137 patients (80%) who were seizure-free at the 6-month post-operative examination. The 2-year post-operative examination was returned by 161 patients, of whom 114 (71%) were seizure-free. There were 88 patients who returned the questionnaire sent 3 years after the surgery, and 58 of them (66%) reported that they were seizure-free. There were 71 patients who returned the questionnaire sent 5 years after surgery; 41 of them (58%) reported that they were seizure-free.

At 6-month and 2-year post-operative examinations, all but one patient received AED. At the 3-year examination, 8 out of 88 (9%) patients received no AED, while at the 5-year post-operative examination, 23 out of 71 (32%) patients were off AED. All patients who stopped taking AED were seizure-free.

Patients lost to follow-up

Instead of the expected 184 patients, only 171 (93%) patients returned to the 6-month post-operative examination. Of the expected 173 returnees, 161 (93%) patients returned to the 2-year post-operative examination. Instead of the expected 95 patients, 88 (92%) patients sent back the questionnaire 3 years after surgery and, of the expected 82, 71 (86%) patients sent back the questionnaire 5 years post-operatively.

During the follow-up period, two patients died from sudden unexpected epileptic death and one patient committed suicide. The other reasons for losing patients to follow-up remain unknown. At each follow-up time point, we compared the clinical data of patients who returned to examinations or sent back the questionnaire with the data of those who were lost to follow-up. At each follow-up time point, there were no significant differences in these two groups with regard to age, epilepsy duration or presence of SGTCS at any follow-up examinations (Mann–Whitney U and Fisher's exact tests were used for these comparisons).

Univariate assessment of prediction

Tables 14 show the results of univariate logistic regression analyses identifying variables which were associated with the outcome at different evaluation times. No factors were significantly related to the short-term (6-month) outcome (Table 1). SGTCS and ictal dystonia were associated with worse 2-year outcome (Table 2). Both these variables together with older age and longer epilepsy duration were also related to a worse 3-year outcome (Table 3). Ictal limb dystonia, older age and longer epilepsy duration were associated with long-term surgical failure evaluated 5 years post-operatively (Table 4). Conclusively, ictal limb dystonia seemed to be the most consistent negative predictor. SGTCS were related to worse outcome 2 and 3 years after the operation, while age and the duration showed an association with long-term outcome. Thus, some variables predicted the long-term outcome, while others only the short-term outcome.

View this table:
Table 1

Univariate analysis of the variables at 6-month post-operative evaluation

Mean (SD) in seizure-free patients 6 months after surgeryMean (SD) in non-seizure-free patients 6 months after surgeryOR (95% CI)
Age (years)*32.3 (10.1)36.6 (11.2)0.96 (0.93–1.00)
Age at epilepsy onset (years)*10.4 (7.8)13.1 (8.4)0.96 (0.92–1.01)
Epilepsy duration (years)*22.0 (10.8)23.5 (9.9)0.99 (0.95–1.02)
Number (%) in seizure-free patientsNumber (%) in non-seizure-free patientsOR (95% CI)
Males54 (39.4)17 (50.0)0.65 (0.31–1.38)
History of febrile seizures72 (52.6)13 (38.2)1.79 (0.83–3.86)
History of status epilepticus14 (10.3)2 (5.9)1.83 (0.40–8.48)
Presence of SGTCS83 (60.6)25 (73.5)0.55 (0.24–1.28)
Presence of frequent SGTCS (at least one per month)18 (13.1)6 (17.6)0.71 (0.26–1.94)
>4 seizures/month71 (51.8)21 (61.8)0.67 (0.31–1.44)
Ictal limb dystonia51 (37.5)17 (50.0)0.60 (0.28–1.28)
Postictal aphasia49 (36.0)7 (20.6)2.17 (0.88–5.35)
Unilateral IED101 (74.8)20 (64.5)1.63 (0.71–3.75)
Contralateral seizure pattern7 (5.1)3 (8.8)0.56 (0.14–2.29)
‘Standard’ temporal lobectomy80 (58.4)24 (70.6)0.59 (0.26–1.32)
  • * In case of continuous variables, the OR and CI are related to 1 SD elevation.

View this table:
Table 2

Univariate analysis of the variables at 2-year post-operative evaluation

Mean (SD) in seizure-free patients 2 years after surgeryMean (SD) in non-seizure-free patients 2 years after surgeryOR (95% CI)
Age (years)*31.7 (10.0)34.9 (10.7)0.97 (0.94–1.00)
Age at epilepsy onset (years)*10.9 (8.1)10.4 (61.1)1.01 (0.96–1.06)
Epilepsy duration (years)*20.9 (10.7)24.5 (9.8)0.97 (0.94–1.00)
Number (%) in seizure-free patientsNumber (%) in non-seizure-free patientsOR (95% CI)
Males48 (42.1)19 (40.4)1.07 (0.54–2.14)
History of febrile seizures58 (50.9)22 (46.8)1.18 (0.60–2.32)
History of status epilepticus11 (9.7)5 (10.6)0.91 (0.30–2.77)
Presence of SGTCS65 (57.0)37 (78.7)0.36 (0.16–0.79)
Presence of frequent SGTCS (at least one per month)14 (12.3)6 (12.8)0.96 (0.34–2.66)
>4 seizures/month61 (53.5)28 (59.6)0.78 (0.39–1.57)
Ictal limb dystonia40 (35.4)25 (53.2)0.48 (0.24–0.96)
Postictal aphasia39 (34.5)15 (31.9)1.12 (0.54–2.32)
Unilateral IED84 (75)31 (68.9)1.36 (0.63–2.90)
Contralateral seizure pattern7 (6.2)3 (6.4)0.97 (0.24–3.92)
‘Standard’ temporal lobectomy63 (55.3)33 (70.2)0.52 (0.25–1.08)
  • * In case of continuous variables, the OR and CI are related to 1 SD elevation.

  • The figures in bold indicate a statistically significant association.

View this table:
Table 3

Univariate analysis of the variables at 3-year post-operative evaluation

Mean (SD) in seizure-free patients 3 years after surgeryMean (SD) in non-seizure-free patients 3 years after surgeryOR (95% CI)
Age (years)*29.9 (9.8)35.3 (9.6)0.94 (0.90–0.99)
Age at epilepsy onset (years)*11.5 (8.4)9.9 (6.2)1.03 (0.97–1.09)
Epilepsy duration (years)*18.4 (10.2)25.4 (10.2)0.94 (0.90–0.98)
Number (%) in seizure-free patientsNumber (%) in non-seizure-free patientsOR (95%CI)
Males23 (39.7)16 (53.3)0.58 (0.24–1.40)
History of febrile seizures31 (53.4)15 (50.0)1.15 (0.48–2.77)
History of status epilepticus9 (15.5)3 (10.0)1.65 (0.41–6.63)
Presence of SGTCS34 (58.6)25 (83.3)0.28 (0.10–0.85)
Presence of frequent SGTCS (at least one per month)7 (12.1)6 (20)0.55 (0.17–1.81)
>4 seizures/month31 (53.4)21 (70.0)0.49 (0.19–1.26)
Ictal limb dystonia18 (31.0)19 (63.3)0.26 (0.10–0.66)
Postictal aphasia24 (41.4)8 (26.7)1.94 (0.74–5.09)
Unilateral IED42 (72.4)19 (65.5)1.38 (0.53–3.60)
Contralateral seizure pattern3 (5.2)3 (10.0)0.49 (0.09–2.60)
‘Standard’ temporal lobectomy37 (63.8)22 (73.3)0.64 (0.24–1.69)
  • * In case of continuous variables, the OR and CI are related to 1 SD elevation. The figures in bold indicate a statistically significant association.

View this table:
Table 4

Univariate analysis of the variables at 5-year post-operative evaluation

Mean (SD) in seizure-free patients 5 years after surgeryMean (SD) in non-seizure-free patients 5 years after surgeryOR (95% CI)
Age (years)*29.1 (9.7)35.6 (9.1)0.93 (0.88–0.98)
Age at epilepsy onset (years)*11.5 (8.3)9.6 (5.5)1.04 (0.97–1.11)
Epilepsy duration (years)*17.6 (9.9)26.0 (10.4)0.92 (0.88–0.97)
Number (%) in seizure-free patientsNumber (%) in non-seizure-free patientsOR (95% CI)
Males18 (43.9)14 (46.7)0.89 (0.35–2.30)
History of febrile seizures20 (48.8)20 (66.7)0.48 (0.18–1.26)
History of status epilepticus9 (22)1 (3.3)8.16 (0.97–68.37)
Presence of SGTCS26 (63.4)25 (83.3)0.35 (0.11–1.10)
Presence of frequent SGTCS (at least one per month)4 (9.8)7 (23.3)0.36 (0.09–1.35)
>4 seizures/month22 (53.7)19 (63.3)1.04 (0.37–2.90)
Ictal limb dystonia12 (29.3)16 (53.3)0.36 (0.14–0.97)
Postictal aphasia16 (39.0)7 (23.3)2.10 (0.73–6.03)
Unilateral IED29 (70.7)21 (70)1.04 (0.37–2.90)
Contralateral seizure pattern3 (7.3)2 (6.7)1.11 (0.17–7.06)
‘Standard’ temporal lobectomy27 (65.9)21 (70.0)0.83 (0.30–2.28)
  • * In case of continuous variables, the OR and CI are related to 1 SD elevation. The figures in bold indicate a statistically significant association.

Multivariate analyses

We performed forward selection analyses in order to determine the independent predictors of outcomes at the four evaluation times. We found no independent predictors for the 6-month outcome. The presence of SGTCS [P = 0.022; OR for seizure-free outcome = 0.39; 95% confidence interval for OR (95% CI) = 0.17–0.87] and ictal limb dystonia (P = 0.039; OR = 0.47; 95% CI = 0.23–0.96) independently predicted the 2-year outcome. Longer epilepsy duration (P = 0.014; OR = 0.94; 95% CI = 0.89–0.98) and ictal limb dystonia (P = 0.016; OR = 0.3; 95% CI = 0.11–0.8) independently predicted the 3-year outcome. Longer epilepsy duration (P = 0.003; OR = 0.92; 95% CI = 0.87–0.97) predicted a poor result after 5 years. ROC analysis revealed that epilepsy duration is a reliable predictor for a poor 5-year outcome. The area under the ROC curve was found to be 0.73 (Fig. 1). Fig. 2 demonstrates the association between epilepsy duration and the long-term outcome.

Fig. 1

ROC curve for epilepsy duration as predictor of a poor 5-year outcome. Sensitivity and 1 minus specificity (ROC curves) are shown at various combinations of duration for predicting surgical outcome. A perfect prediction would have 100% sensitivity and 100% specificity, and would include a point at the upper left-hand corner. The curve for a variable with no predictive value would appear as a diagonal line from the lower left to the upper right corner. The area under the curve was 0.73, suggesting that epilepsy duration is a reliable predictor for the 5-year outcome.

Fig. 2

Association of epilepsy duration and 5-year surgical outcome.

Analyses with repeated logistic models revealed that epilepsy duration showed significant differences in predicting the 6-month versus 5-year outcome (P = 0.013), and a non-significant trend 2-year versus 5-year outcome (P = 0.057). Epilepsy duration had a significantly higher prognostic value in predicting the 5-year outcome compared with earlier follow-up. Fig. 3 shows the mean epilepsy duration in seizure-free versus non-seizure-free patients at different evaluation times.

Fig. 3

Mean epilepsy duration in seizure-free versus non-seizure-free patients at different evaluation times. At the 6-month outcome, the mean epilepsy duration was almost equal in seizure-free versus non-seizure-free patients. The difference for epilepsy duration between the seizure-free and non-seizure-free groups became gradually larger at later evaluation times.

It was surprising that unilateral IED showed no association with the outcome in the present study. One of the reasons might be that, after the unilateral irritative zone was reported as a strong positive prognostic factor, some patients with bilateral IED did not undergo an operation. To investigate this question, we examined the 230 patients who met all the inclusion criteria for this study with exception of the operation and follow-up criteria (such as all patients were included who underwent a presurgical evaluation, aged >16 years, had complex partial seizures, had HS on the MRI, but temporal lobectomy or adequate follow-up was not obligatory). Among them, 37 did not have operations. Unilateral IED were present in 53% of non-operated patients and 72% of operated patients (P = 0.036).

Discussion

Determining prognostic factors for the long-term outcome may have methodological pitfalls. One of these pitfalls is that the post-operative outcome changes over time. Many initially seizure-free patients may have had deterioration (Berkovic et al., 1995; Wieser et al., 2003; Yoon et al., 2003), while some initially non-seizure-free patients may achieve long-term seizure freedom (Rasmussen, 1970; Ficker et al., 1999; Wieser et al., 2003). The latter is called ‘running down phenomenon’ and used to refer to the phenomenon of late remission of postsurgical seizures. This occurs in 5–20% of TLE surgery cases (Rasmussen, 1970; Bladin, 1987; Berkovic et al., 1995; Ficker et al., 1999; Wieser et al., 2003). Consequently, studies that consider patients having post-operative seizures at any time post-operatively to have non-seizure-free outcome regardless of any ongoing changes in seizure status ignore the possibility of long-term improvement (e.g. those studies which performed Kaplan-Meier survival analysis; Foldvary et al., 2000). This approach may lead to incorrect conclusions by miscategorizing patients that have experienced seizures initially, but later become seizure-free. Another type of study considered patients seizure-free if they attained 12 consecutive months of absolute seizure freedom regardless of the seizure situation at the last post-operative examination (Hennessy et al., 2000b). This approach does not consider that many initially seizure-free patients may deteriorate. A third type of study used cross-sectional outcome measurement: they considered only the outcome of the last post-operative examinations. Again, this approach is theoretically problematic because the outcome changes over time and mixing the patients who experienced 5-year seizure freedom with those who had only 1-year seizure freedom may result in incorrect conclusions. Short-term and long-term prognosis may have different prognostic factors. Consequently, the cross-sectional outcome studies may be inadequate to study the factors influencing the long-term outcome. These methodological problems might explain the wide variety of predictive factors found by previous studies (McIntosh et al., 2001).

In the present study, we aimed to identify pre-operative predictive factors which influence the short- and long-term outcome in TLE-HS—the most frequent surgically treated epilepsy syndrome with the most frequent underlying pathology (Wieser et al., 2004). We found that ictal dystonia was associated with worse outcome 2 and 3 years after the operation; SGTCS was a negative predictor for the 2-year outcome. Longer epilepsy duration predicted poor long-term outcome. Figure 4 demonstrates the factors that influenced the long-term and short-term outcomes. In the case of epilepsy duration, the prediction value regarding 6-month versus 5-year outcome was found to be significantly different. The difference in predictive powers of pre-operative factors in short versus long-term outcome may explain why studies using various methods of outcome assessments (such as variety in minimal follow-up times, cross-sectional, or survival analysis) (McIntosh et al., 2001) found various predictors.

Fig. 4

Predictors for long-term and short-term outcomes. Longer epilepsy duration seems to be the most important negative predictor for the long-term results of surgery of TLE-HS. Only those variables are presented which were found to be significant in both univariate and multivariate analyses. *Analyses with repeated logistic models revealed that epilepsy duration showed significant differences in predicting the 6-month versus 5-year outcome (P = 0.013). The other variables showed no significant differences in these comparisons; thus, we cannot state that the prognostic value of ictal dystonia and presence of SGTCS changes over time post-operatively, but they were prognostic factors exclusively for 2-year and 3-year outcome, respectively.

There are only three reports of homogeneous TLE patient cohorts with HS and, in all of them, HS was defined pathologically and post-operatively (Kilpatrick et al., 1999; Hennessy et al., 2000b; Hardy et al., 2003). Kilpatrick et al. (1999) restricted the analysis to the historical features of 56 operated patients using cross-sectional outcome assessment (minimum follow-up was 20 months) and concluded that age at operation, epilepsy duration, or presence of SGTCS were not predictive factors. Conversely, our study found that epilepsy duration and presence of SGTCS are predictors for outcome. However, the epilepsy duration predicted only the long-term, while SGTCS only the 2-year outcome. Hennessy et al. (2000b) investigated the probability of achieving at least 1 year of seizure-freedom in patients who underwent temporal lobectomy and who had HS on the post-operative pathological analysis. They found that the factor contributing to favourable outcome was the presence of unilateral IED, while perinatal complications and SGTCS were associated with worse outcome. We also found that SGTCS were associated with worse outcome. Concerning perinatal complications, in our study population there were only five patients who definitely had perinatal complications. These data are not presented here and, due to limited statistical power, we did not select this variable for analysis. Hardy et al. (2003) investigated patients who underwent temporal lobectomy and who had HS at the post-operative pathological examination found by cross-sectional outcome assessment with 1-year minimum follow-up. They found that the history of status epilepticus predicted surgical failure. Compared with our results, we did not find a significant association of the history of status epilepticus with the outcome.

Some previous studies showed that a long duration of epilepsy is associated with poorer outcome after surgery (Guldvog et al., 1994; Eliashiv et al., 1997; Specht et al., 1997; Morris et al., 1998; Jeong et al., 1999; Hennessy et al., 2001; Wieser et al., 2003). Other studies did not find such an association (Blume et al., 1994; Radhakrishnan et al., 1998; Kilpatrick et al., 1999; Janszky et al., 2000). A recent study (Yoon et al., 2003) examining patients with different epilepsy syndromes and underlying aetiologies analysed the factors which were responsible for the relapse after an initial post-operative seizure-free period. This study found that the duration of epilepsy >20 years and normal pathology predicted the late relapse. In accordance with that study, we found that duration of epilepsy is not predictive for the short-term outcome, but highly predictive for the long-term outcome.

The mechanism whereby chronicity of seizures imparts poor outcome is uncertain, but this may suggest that secondary epileptogenesis at sites distant to the lesion may develop with years of uncontrolled seizures. TLE-HS seems to be a progressive disorder. After a childhood initial precipitating injury (e.g. febrile seizures), the first unprovoked seizure appears only some years later after a silent period (French et al., 1993; Wieser et al., 2004). After the first unprovoked seizure, it takes some additional years for epilepsy to become pharmacologically resistant (Berg et al., 2003; Wieser et al., 2004). Now we provide evidence that, after pharmacological intractability, it takes further years for this epilepsy to become surgically intractable.

Longer epilepsy duration associates with chronic structural and functional abnormalities. Jokeit et al. (1999) found an association between the epilepsy duration and the bilateral decline of hippocampal volume, brain glucose metabolism and Wada hemispheric memory performance. Contralateral epileptogenesis, however, can only partly be responsible for the long-term surgical failure. Hennessy et al. (2000a) analysed seizures of seven TLE-HS patients who each had a relapse after an initial seizure-free post-operative period. The post-operative seizures in four patients originated from the ipsilateral and in three patients from the contralateral hemisphere.

Like our study, other studies also found that the presence of SGTCS was a poor predictive factor for some previous outcome studies investigating TLE (Bengzon et al., 1968; Blume et al., 1994; Specht et al., 1997; Hennessy et al., 2000b) or frontal lobe epilepsy (Janszky et al., 2000). Some other studies did not find such a relationship (Dodrill et al., 1986; Guldvog et al., 1994). This suggests that patients with SGTCS may have an extended epileptogenic region compared with patients who had only focal seizures. SGTCS are associated with multifocal irritative areas (Blume, 1978). TLE patients who had SGTCS prior to PET examination have an extended hypometabolism compared with those who have no SGTCS (Savic et al., 1997).

Presence of ictal lateralizing signs was found to be a positive predictor for surgery of posterior epilepsy (Boesebeck et al., 2002). Previous outcome studies of TLE surgery did not include variables from video-EEG evaluation, although this investigation is one of the key elements of modern presurgical evaluation (Rosenow and Lüders, 2001). Thus, in this study we chose two well-defined peri-ictal lateralizing signs defined by video-EEG, namely the postictal aphasia (Gabr et al., 1989) and ictal dystonia (Kotagal et al., 1989). Both of them are characteristic features of TLE-HS (Wieser et al., 2004). Postictal aphasia was not found to be a predictor, while the ictal dystonia was associated with worse outcome at 2-year and 3-year evaluation times. This was surprising because ictal dystonia is a characteristic feature of mesial TLE; it distinguishes between lateral and mesial TLE (Dupont et al., 1999; Pfander et al., 2002). Conversely, since all of our patients had mesial TLE, ictal dystonia may not be associated with a more circumscribed epileptogenic region within the mesial TLE spectrum. Ictal dystonia is accompanied by widespread activation of the temporal and frontal lobes according to an invasive video-EEG study (Kuba et al., 2003) or with a seizure propagation to basal ganglia according to a SPECT study (Newton et al., 1992). Our previous study found that spreading of ictal activity is accompanied by worse outcome (Schulz et al., 2000).

It was surprising that unilateral IED, which had been so consistently found as a predictor by previous studies, showed no association with the outcome in the present study. However, there are other studies which also failed to demonstrate the predictive value of unilateral spike focus (Jeong et al., 1999; Hennessy et al., 2001; Clusmann et al., 2002). One of the reasons might be that, after the unilateral irritative zone was reported as a strong positive prognostic factor (Holmes et al., 1997; Radhakrishnan et al., 1998; Schulz et al., 2000), we predicted reduced likelihood for seizure freedom in counselling our TLE patients with bilateral IED. Consequently, some patients might have refused the operation for this reason. Indeed, we found that unilateral IED occurred significantly more often in the operated patients than in those patients who underwent presurgical evaluation but finally were not operated on, suggesting that patients with unilateral IED are more likely to be chosen for operation. An alternative explanation would be that, in surgery of TLE-HS, many previously well-identified predictors for surgery of TLE due to various aetiologies lose their predictive value (Hardy et al., 2003).

Some studies found that TLE together with a history of febrile seizures is associated with favourable outcome (Abou-Khalil et al., 1993; Kanemoto et al., 1998). These early studies ignored, however, whether the patients had HS. Because HS is often associated with a history of febrile seizures, it is reasonable to assume that, in these patients, the favourable outcome might be related to the HS and not to the febrile seizures since TLE-HS has excellent surgical outcome (Radhakrishnan et al., 1998). In our recent study (Janszky et al., 2003), we investigated whether the mesial TLE coupled with complex febrile seizures has different clinical features and surgical prognosis in comparison with the medial TLE without a history of febrile convulsions. In that study, we excluded patients with simple febrile seizures or those in whom it was not documented whether they had had simple or complex febrile convulsions. We found that the outcome after temporal lobe surgery was significantly more favourable in patients with a history of complex febrile seizures than in patients who had no febrile seizures. In the present study, however, we did not find febrile seizures to be predictive, and we did not make a distinction between complex versus simple febrile seizures because in some cases it is impossible to abstract this information from medical records.

AED withdrawal did not seem to affect the 3-year and 5-year outcome because, at 5-year outcome assessment, 32% of patients were seizure-free without receiving AED, while all patients who were non-seizure-free never had a complete AED withdrawal during the evaluation. In our protocol, AED reduction begins only 2 years after the surgery and takes months or even years to reach complete withdrawal. In this study, however, we did not consider whether a partial discontinuation or reduction of AED is associated with a seizure recurrence, which is one of the limitations of our study.

Conclusively, we found that the predictors for the long-term surgical outcome of TLE with HS are different from those variables that predict the short-term outcome. Epilepsy duration seems to be the most important predictor for the long-term results of surgery of TLE-HS. Our results strongly suggest that surgery for TLE-HS should be performed earlier than presently done (Trevathan and Gilliam, 2003). TLE-HS has poor pharmacological (Semah et al., 1998) and excellent surgical prognosis (Radhakrishnan et al., 1998), but we suggest that these excellent chances gradually decrease with longer epilepsy duration (Fig. 3). Furthermore, need for an early operation is required not only due to the seizure outcome but also in order to prevent increasing social burden, growing neuropsychological deficit and psychiatric co-morbidity associated with longer epilepsy duration (Engel, 1999; Oyegbile et al., 2004).

Acknowledgments

We wish to thank Hermann Steffen who helped us in data collection and organized the post-operative examinations. We also wish to thank Terri Shore Ebner who reviewed the manuscript as a native English speaker. This work was supported by the Humboldt Foundation and a grant from the Deutsche Forschungsgemeinschaft (DFG-Eb 111/2–2). Our work was also supported by the Ansgarius Foundation, Stockholm.

References

View Abstract