Factors Predicting Outcome of Surgery for Intractable Epilepsy with Pathologically Verified Mesial Temporal Sclerosis


Address correspondence and reprint requests to Dr. J. W. Miller at Regional Epilepsy Center (Box 359745), Harborview Medical Center, Seattle, WA 98104-2499, U.S.A. E-mail: millerjw@u.washington.edu


Summary:  Purpose: To examine the subgroup of patients with medically intractable epilepsy receiving temporal lobectomies who have pathologically verified mesial temporal sclerosis (MTS) and to determine the relation of demographic and clinical factors, results of diagnostic testing, and details of the surgical procedure with prognosis for achieving control of seizures.

Methods: All patients receiving surgical treatment for intractable epilepsy between 1991 and 1998 at the University of Washington were reviewed. There were 118 patients who met inclusion criteria of adequate pathological analysis showing MTS without a progressive process and a minimum of 1-year follow-up.

Results: Only personal history of status epilepticus demonstrated significant (p = 0.0276) prediction of outcome, increasing the risk of surgical failure. No other factors were significant predictors of outcome, including history of febrile seizures, possible etiologic factors, EEG, magnetic resonance imaging (MRI) or neuropsychological testing results, or extent of resection.

Conclusions: Many factors that have been previously described to predict favorable outcome in the overall group of patients receiving temporal lobe resections for intractable epilepsy are, in fact, predictors of MTS and lose their predictive value when the subgroup of patients with confirmed MTS is examined. Neurosurgical treatment of MTS can be very effective even in the presence of significant etiologic factors, or of bilateral or extratemporal abnormalities on EEG or MRI.

Mesial temporal sclerosis (MTS) is the most common pathological finding in resective surgery for temporal lobe epilepsy (1–3). Some clinical predictors of favorable outcome for temporal lobe resections are associated with the presence of MTS. These predictors include a history of febrile seizures (4), unilateral decrease in hippocampal volume, and other changes on magnetic resonance imaging consistent with MTS (5). Although MTS clearly indicates a good surgical prognosis, little attention has been paid to the predictors of surgical outcome within the subgroup of patients with pathologically proven MTS, to try to explain why temporal resection does not always lead to seizure control in these patients. To address this, we reviewed the entire surgical experience of the University of Washington Regional Epilepsy Center over a 7-year period to identify all patients who met criteria of pathologically verified MTS and had adequate follow-up to determine retrospectively the association of demographic features, diagnostic testing, and details of the surgical procedure with outcome.


All patients (N = 492) who received surgical treatment for intractable seizures at the University of Washington Regional Epilepsy Center between 1991 and 1998 were reviewed for possible inclusion in the study. Those patients who had adequate tissue analysis that showed the well-described (6) criteria of MTS (n = 147) were selected by an investigator (D.E.B.) who was blinded to the clinical details of the patient's condition. Inclusion required that two of three cell regions (CA1, CA3, endfolium) were available to review and showed severe neuron loss (qualitatively >50%). Patients with evidence of a progressive process, specifically a malignant or unresectable neoplasm, were excluded, leaving 142 patients. After efforts to maximize follow-up information with medical record review and telephone interviews, 118 patients met criteria for a minimum of 1-year follow-up information on seizure occurrence after surgery. The telephone script included questions to determine seizure frequency during the past year relative to the preoperative seizure rate. Procedures for data collection and the script for telephone interviews were approved by the University of Washington Human Studies Committee.

Demographic data, results of diagnostic testing, and details of the surgical resection were obtained from three sources: (a) data recorded in a database at the time of presentation of the patients at a multidisciplinary presurgical conference before surgery (kept by C.B.D.); (b) data recorded by the neurosurgeon (G.A.O.) before and after surgery, and (c) review of source documents in the medical records.

Nearly all patients completed a standard presurgical evaluation consisting of long-term video-EEG monitoring; magnetic resonance imaging (MRI) with high-resolution phased-array views of the temporal lobes; a consistent, standardized neuropsychology battery (7); and an intracarotid amobarbital procedure. A small number of patients had some of these studies performed at outside institutions, which were included in our data analysis, if the studies were reviewed at the University of Washington. MRI studies were reviewed by a neuroradiologist (D.H.) whenever possible to confirm the findings, but as the imaging studies were performed over a 7-year period with variation in technique, volumetric evaluation of hippocampal structures was not attempted. Of these studies, 77% included high-resolution temporal lobe evaluation with phased-array surface coils, with the coronal imaging plane oblique to the long axis of the hippocampus. The high-resolution evaluation included a T2 fast spin-echo (FSE) examination with additional sequences consisting of white-matter inversion recovery, fluid-attenuated inversion recovery (FLAIR), proton density, or a combination thereof. For the T2 FSE sequence, a matrix of 512 × 384 with field of view (FOV) of 18 cm was used, with lower in-plane resolution for the other sequences. Four-millimeter slice thickness was used. Patients not evaluated with surface coils were scanned in head coil only, with most undergoing oblique coronal imaging with proton density and T2 with FSE or conventional spin-echo sequences. In these cases, a 256 × 256 matrix with 5-mm slice thickness was used. Four patients were evaluated with axial proton density and T2 conventional imaging only, and three were judged to exhibit findings of MTS.

All resections were performed by a single neurosurgeon (G.A.O.), who made consistent measurements of resection parameters during surgery. The extent of the lateral temporal and hippocampal resection was tailored by using intraoperative electrocorticography, with surgical procedures previously described (8).

Patients were categorized into one of three groupings, with criteria that have been in use at our center for many years (8): class I, seizure free, except auras; class II, worthwhile improvement with a ≥75% reduction of seizure occurrence; and class III, <75% reduction of seizures. Outcome data were obtained from a neurosurgical database (GAO), the medical records, and telephone interviews.


Improvement in seizure control after surgery occurred in 92.3% of the 118 patients who met criteria for inclusion. Of these, 73 (61.8%) were seizure free (class I), 36 (30.5%) were in class II, and nine (7.7%) were in class III.

Demographic and clinical factors

Demographic characteristics of the patients—age, gender, and handedness—were similar among the different outcome groups (Table 1). Length of follow-up and region of origin were similar among the outcome groups. A history of status epilepticus was a significant predictor of poor surgical outcome, but no other item in the neurologic history predicted outcome. Significant etiologic factors included histories of significant head trauma (defined as trauma-associated loss of consciousness for >24 h, skull fracture, or intracranial hemorrhage), tuberous sclerosis, presence of ventriculoperitoneal shunt, arteriovenous malformation, CNS infection, global hypoxia, or infantile hemiparesis. Surprisingly, presence of such etiologic factors was not associated with a poorer prognosis. Abnormal neurologic examination also did not predict outcome.

Table 1.  Clinical characteristics and seizure outcome
Clinical featureTotalClass IClass IIClass IIIp value
  • Percentage of category in parentheses; ± indicates standard deviation.

  • a

     Not significant (NS), Bonferroni-corrected Student's t test.

  • b

     Fisher's exact test.

Age at seizure onset (yr)8.8 ± 9.48.5 ± 8.910.0 ± 11.04.4 ± 4.5NSa
Age at surgery (yr)31.5 ± 10.432.1 ± 10.632.0 ± 9.324.4 ± 10.8NSa
GenderMale, 5130 (58.8)18 (35.3)3 (5.9)0.5599b
 Female, 6743 (64.2)18 (26.9)6 (9.0) 
HandednessRight, 9357 (61.3)28 (30.1)8 (8.6)0.8823b
 Left, 2314 (60.9)8 (34.8)1 (4.3) 
 Mixed or unclear, 22 (100.0)0 (0.0)0 (0.0) 
Follow-up period (yr)3.8 ± 2.23.4 ± 2.44.4 ± 2.85.1 ± 2.2NSa
Region of originSeattle metro, 4224 (57.1)15 (35.7)3 (7.1)0.3372b
 Other Washington State, 4827 (56.3)16 (33.3)5 (10.4) 
 Outside Washington State, 2822 (78.6)5 (17.9)1 (3.6) 
Family history of seizuresYes, 2918 (62.1)9 (31.0)2 (6.9)NSb
 No, 8955 (61.8)27 (30.3)7 (7.9) 
History of febrile seizuresYes, 3021 (70.0)8 (26.7)1 (3.3)NSb
 No, 8852 (59.1)28 (31.8)8 (9.1) 
History of status epilepticusYes, 158 (53.3)3 (20.0)4 (26.7)0.0276b
 No, 10365 (63.1)33 (32.0)5 (4.9) 
Psychiatric historyYes, 3217 (53.1)10 (31.3)5 (15.6)0.1326b
 No, 8656 (65.1)26 (30.2)4 (4.7) 
Significant causal factorsYes, 3722 (59.5)12 (32.4)3 (8.1)0.9511b
 No, 8151 (63.0)24 (29.6)6 (7.4) 
Neurologic examinationNormal, 9258 (63.0)26 (28.3)8 (8.7)0.5285b
 Abnormal, 2615 (57.7)10 (38.5)1 (3.8) 

Diagnostic testing

Findings on EEG recordings did not correlate with surgical outcome (Table 2), as rates of seizure control were similar with different patterns of interictal and ictal abnormalities. Even the decision to perform invasive EEG monitoring (eight subdural grids with additional subdural strips, 20 subdural strips only) did not correlate with a worse outcome. A possible exception might be the extremely small subgroup of four patients who had more than one region of ictal onset, as only one of these became seizure free. Interictal spikes were present in all 118 patients. Preoperative neuropsychological testing, including Verbal, Performance, and Full Scale IQ scores on the Wechsler Adult Intelligence Scale–Revised did not correlate with outcome. MRI findings also did not predict outcome. The typical MRI findings of MTS were not demonstrated in 20 patients, even though they were later found to have MTS on pathological examination, indicating that mild MTS can occasionally be missed with the imaging methods used during the period of this study. Of these 20 patients, 16 had surgery prior to 1993.

Table 2.  Diagnostic tests
Diagnostic testTotalClass IClass IIClass IIIp value
  • Percentage of category in parentheses; ± indicates standard deviation.

  • a

     Fisher's exact test.

  • b

     Not significant (NS), Bonferroni-corrected Student's t test.

EEG: Interictal epileptiform abnormalityIpsilateral temporal only, 6542 (64.6)18 (27.7)5 (7.7)0.4844a
 Ipsilateral, temporal + extratemporal, 105 (50.0)3 (30.0)2 (20.0) 
 Bilateral, 4326 (60.5)15 (34.9)2 (4.6) 
EEG: Ictal epileptiform abnormalityIpsilateral temporal only, 9156 (61.5)28 (30.8)7 (7.7)0.3553a
 Additional focus recorded, 41 (25.0)2 (50.0)1 (25.0) 
 No seizures recorded or unclear localization, 2316 (69.6)6 (26.1)1 (4.3) 
EEG: Invasive monitoring neededYes, 2814 (50.0)10 (35.7)4 (14.3)0.4905a
 No, 9059 (65.6)26 (28.9)5 (5.6) 
NeuropsychologyVIQ, 88.3 ± 14.088.1 ± 13.288.0 ± 13.091.1 ± 22.3NSb
 PIQ, 88.5 ± 13.489.4 ± 13.487.3 ± 11.685.6 ± 19.0NSb
 FSIQ, 87.4 ± 13.787.6 ± 13.386.9 ± 12.388.1 ± 21.4NSb
MRIUnilateral hippocampal atrophy only, 7248 (66.7)20 (27.8)4 (5.6)0.3855a
 Additional abnormalities, 2616 (61.5)8 (30.8)2 (7.7) 
 Hippocampal, 74 (57.1)1 (14.3)2 (28.6) 
 Extrahippocampal, 1912 (63.2)7 (36.8)0 (0.0) 
 Normal or unclear, 209 (45.0)8 (40.0)3 (15.0) 

Surgical details

Resections on the left side, or in the speech-related hemisphere, had somewhat better outcomes (Table 3), but this difference was not statistically significant. Little difference was found in the extent of surgical resection among the class I, II, and III outcome groups.

Table 3.  Details of surgery
Surgical detailTotalClass IClass IIClass IIIp
  • Percentage of category in parentheses; ± indicates standard deviation.

  • a

     Fisher's exact test.

  • b

     Not significant (NS), Bonferroni-corrected Student's t test.

Side of resectionRight, 5832 (55.2)20 (34.5)6 (10.3)0.3158a
 Left, 6041 (68.3)16 (26.7)3 (5.0) 
 Nondominant, 6133 (54.1)21 (34.4)7 (11.5)0.3661a
 Dominant, 5034 (68.0)14 (28.0)2 (4.0) 
 Mixed or unclear dominance, 76 (85.7)1 (14.3)0 (0.0) 
Extent of resection (mm)Medial temporal, 62.9 ± 9.561.7 ± 9.565.5 ± 9.263.4 ± 9.7NSb
 Lateral temporal, 47.8 ± 15.147.9 ± 14.447.2 ± 14.949.6 ± 23.0NSb
 Hippocampus, 29.3 ± 7.028.5 ± 6.931.0 ± 7.629.9 ± 3.8NSb


Only history of status epilepticus was a significant predictor of postoperative outcome. It is likely that this history is associated with a greater chance of a more extensive or multifocal epileptogenic zone, which may not adequately respond to a standard temporal lobe resection. Remarkably, no other factors examined were found to have definite prognostic power in this group of patients with pathologically proven MTS. However, because surgery failed for only nine patients, it is possible that with a much larger patient population, additional predictors could be discovered.

As might be expected, these results contrast substantially with numerous past studies of the overall group of patients receiving temporal lobe resections for epilepsy, which have identified several predictors of outcome. The best documentation is for favorable outcome with unilateral hippocampal atrophy on MRI (3,5,9,11,12), and concordant interictal epileptiform abnormalities (4,5,9,10). Other factors that have been shown to predict outcome in these series include duration of epilepsy (3,13), age at operation (3,4,13), history of febrile seizures (4), and presence of a known etiology for epilepsy (2).

However, the findings of the present study are strikingly similar to those of the two previously published series of patients with pathologically verified MTS (14,15). Those studies failed to show that age at onset and operation, duration of epilepsy, or history of febrile seizures affected outcome (14,15). The largest prior study, by Hennessy et al. (14), is similar in size (116 patients), duration of follow-up, and overall outcome to the present study, but covered a substantially longer and earlier period, 1975 to 1995. That study did not examine a history of status epilepticus, but did note that a history of secondarily generalized seizures and interictal spikes not confined to the ipsilateral temporal lobe did predict a worse outcome. That study differs from ours in that six patients with dual pathology were excluded; many of the patients did not receive MRI; and the majority had only interictal EEG data without long-term video-EEG monitoring. In that study, 78% of operated-on patients had interictal spikes confined to the ipsilateral temporal lobe, as compared with only 55% in the current study. Modern neuroimaging and EEG monitoring techniques have made epilepsy surgery more used for MTS patients with multiple or bilateral interictal spikes, and in the current series, such patients have nearly as good an outcome as do those with interictal spikes confined to the resected temporal lobe (Table 2).

Many factors that predict favorable outcome in the overall group of temporal lobe resections appear to be predictors of MTS, a condition that forecasts a good chance of seizure control (2). Our study demonstrates that if instead one looks at the subgroup of MTS patients, most of these factors lose their predictive value. With modern techniques of presurgical evaluation, neurosurgical treatment of MTS can now be very effective, even in the presence of significant etiologic factors, or of bilateral or extratemporal abnormalities on EEG or MRI. Therefore when preoperative evidence of MTS is found in patients with intractable epilepsy, resective surgery should be strongly considered, even when complex clinical features or neuroimaging findings are present, or invasive EEG monitoring is required.

Acknowledgment: This study was supported in part by the Mary Gates Endowment for Undergraduate Research