The Multicenter Study of Epilepsy Surgery: Recruitment and Selection for Surgery


Address correspondence and reprint requests to Dr. A. T. Berg at BIOS/NIU, Montgomery Hall, DeKalb, IL 60115, U.S.A. E-mail:


Summary: Purpose: Multiple studies have examined predictors of seizure outcomes after epilepsy surgery. Most are single-center series with limited sample size. Little information is available about the selection process for surgery and, in particular, the proportion of patients who ultimately have surgery and the characteristics that identify those who do versus those who do not. Such information is necessary for providing the epidemiologic and clinical context in which epilepsy surgery is currently performed in the United States and in other developed countries.

Methods: An observational cohort of 565 surgical candidates was prospectively recruited from June 1996 through January 2001 at six Northeastern and one Midwestern surgical centers. Standardized eligibility criteria and protocol for presurgical evaluations were used at all seven sites.

Results: Three hundred ninety-six (70%) study subjects had resective surgery. Clinical factors such as a well-localized magnetic resonance imaging (MRI) abnormality and consistently localized EEG findings were most strongly associated with having surgery. Of those who underwent intracranial monitoring (189, 34%), 85% went on to have surgery. Race/ethnicity and marital status were marginally associated with having surgery. Age, education, and employment status were not. Demographic factors had little influence over the surgical decision. More than half of the patients had intractable epilepsy for ≥10 years and five or more drugs had failed by the time they initiated their surgical evaluation. During the recruitment period, eight new antiepileptic drugs were approved by the Food and Drug Administration for use in the United States and came into increasing use in this study's surgical candidates. Despite the increased availability of new therapeutic options, the proportion that had surgery each year did not fluctuate significantly from year to year. This suggests that, in this group of patients, the new drugs did not provide a substantial therapeutic benefit.

Conclusions: Up to 30% of patients who undergo presurgical evaluations for resective epilepsy surgery ultimately do not have this form of surgery. This is a group whose needs are not currently met by available therapies and procedures. Lack of clear localizing evidence appears to be the main reason for not having surgery. To the extent that these data can address the question, they suggest that repeated attempts to control intractable epilepsy with new drugs will not result in sustained seizure control, and eligible patients will proceed to surgery eventually. This is consistent with recent arguments to consider surgery earlier rather than later in the course of epilepsy. Postsurgical follow-up of this group will permit a detailed analysis of presurgical factors that predict the best and worst seizure outcomes.

Epilepsy surgery is an acknowledged and accepted treatment for medically refractory seizures (1). Numerous case series have reported significant postsurgical improvements in seizure control with ≤90% of selected patients becoming seizure free (2,3).

Recently, the first successful randomized trial of anterior temporal lobectomy was reported by Wiebe et al. (4). The authors took advantage of the typical 1-year waiting period before patients in Canada can have surgery to perform this study. Patients who agreed to enter the trial were randomized either to have the surgery after the usual 1-year wait (what would typically happen were they not in a trial) or immediately. The results of the study showed a strong positive effect of surgery on seizures as well as on quality of life. Because of the design of the study, the surgical-to-nonsurgical comparison could be continued for only 1 year, at which point the nonsurgical patients tended to have surgery.

Despite this one high-quality trial, much remains unknown regarding specific predictors of good seizure outcome and the relations between the occurrence of seizures and psychosocial outcomes over the long term. In addition, some of the largest studies available are from the pre–magnetic resonance imaging (MRI) era.

In June 1996 we began recruitment of patients for a large multicenter epilepsy surgery study. The intent was to recruit, over a relatively short period, a large cohort of patients who presented for evaluation for resective epilepsy surgery and to describe their presurgical experience, their evaluation and selection for surgery, and their outcomes, including seizures, quality of life, and psychiatric and neuropsychological outcomes. The purpose of this report is to provide a description of the types of patients who present for evaluation for resective surgery, which factors most influence the decision to have surgery, and the proportion of patients who ultimately go on to have surgery.



Six surgical centers in the northeast United States (Yale, New Haven, CT; Jefferson Medical Center, Philadelphia, PA; Columbia University School of Medicine, Montefiore Medical Center, and New York University, New York, NY; and Rochester Medical Center, Rochester, NY) began recruiting patients in June 1996. A seventh center, MINCEP, Minneapolis, MN, was added in December 1997. Recruitment of new patients continued to January 31, 2001. All procedures were approved by the Institutional Review Boards of all participating centers. An appropriate consent procedure was followed with each individual, and signed informed consent obtained for those willing to participate in the study. For those younger than 18 years, signed informed parental permission and signed informed assent of the individual were obtained.


To be eligible for the study, patients had to be ≥12 years old at the time of initial presentation to the center for surgical evaluation. At least two first-line antiepileptic drugs (AEDs) must have failed, and patients had a minimum of 20 partial or secondarily generalized seizures during the previous 2 years as documented by history obtained from the patient and the medical record. We originally specified that trials of at least two of five specified drugs must have failed [carbamazepine (CBZ), phenytoin (PHT), primidone (PRM), phenobarbital (PB), or sodium valproate (VPA)]. Over the period that we recruited patients, the criteria for first-line drug in the community changed because of the approval in the United States of several new AEDs beginning in 1994. Thus some allowance was made for this in our recruitment. We excluded patients who, on evaluation, were found not to meet these basic eligibility criteria, who had only nonepileptic seizures or generalized epilepsy, or whose intellectual functioning was too impaired to complete the necessary study forms. Patients who had had previous epilepsy surgery (but not necessarily previous neurosurgery for other reasons) were also excluded.

Research protocol

Trained research associates at each site monitored all new incoming patients to the epilepsy surgery program. Those who appeared to meet the basic entry criteria were immediately approached and invited to participate in the study. If they agreed, the research associate followed appropriate procedures for obtaining written informed consent from adult participants and written informed permission from parents of minor participants and written informed assent of the minor. Date of study entry for participating patients is therefore essentially the same as date of first being seen for a surgical evaluation, and the terms are used interchangeably. Research associates interviewed participants with the use of a structured questionnaire. Quality of life was measured with the QOLIE-89 (5) for those at least 17 years old. Psychiatric measures included the Beck's Depression and Anxiety indices (BDI and BAI) (6,7). Medical records were reviewed and information coded on a structured form for data entry. All data forms were edited and corrected before data were entered. Records were reviewed and information extracted all the way through the point that the patient had surgery or a decision not to have surgery was made. In this way, the presurgical evaluation process, treatment during that period, and the surgery itself were prospectively recorded in a structured manner. Historical factors before presentation at the center were recorded based on retrospective review but still in a structured, standardized manner. Included in the data collected was information about each AED that a patient had tried before entering the study as well as all recently approved drugs tried during the presurgical evaluation period.

Surgical evaluation protocol

All centers agreed to follow a standard, staged protocol for evaluation of patients. Phase I involved noninvasive testing. This included history, neurologic evaluation, MRI, video-EEG monitoring, and neuropsychological evaluation. Clinical psychiatric evaluations were performed when clinically indicated, and no study criteria or protocols were prescribed. Fluorodeoxyglucose–positron emission tomography (FDG-PET) and single-photon emission computed tomography (SPECT) imaging were performed at those centers in which they were available. Phase II consisted of a cerebral angiogram and a bilateral intracarotid amobarbital procedure (Wada test). Phase III involved invasive (intracranial) EEG monitoring. The criteria for deciding to do intracranial monitoring are presented later. Phase IV is the actual resective surgery with intraoperative functional mapping and electrocorticography as needed.

For the neuropsychological evaluation, all centers also agreed to administer the same core components. These consisted of WAIS or WISC-III depending on age, the California Verbal Learning Test, FAS, Boston Naming Test, Hooper, and the Rey complex figure test. These tests measure a range of cognitive domains typically affected by epilepsy and surgery and are commonly used in pre- and postsurgical evaluations (8).

Based on the results of the neuroimaging studies, patients were divided into those with and without space-occupying lesions. A space-occupying lesion was generally a tumor, a vascular anomaly, or a well-defined developmental abnormality such as a hamartoma or focal cortical dysplasia. Hippocampal atrophy consistent with mesial temporal sclerosis (MTS), other atrophic regions, and diffuse or poorly defined developmental abnormalities were not considered in this category.

A specific protocol was used in patients with space-occupying lesions as defined earlier. Resective surgery could proceed without phase III investigation if an interictal or ictal EEG or both demonstrated abnormal activity in the same area as the space-occupying lesion, and there was no contradictory localizing information from other tests. In these cases, the resection included the lesion and was extended to surrounding cortex to an extent determined by results of intraoperative electrocorticography, pathological evidence of normal cytoarchitectonic margins, or both. If the resection area was adjacent to or in important functional cortex, intraoperative mapping was used. For patients with space-occupying lesions who did not fit these criteria, phase III was done to provide EEG localization as well as extraoperative functional mapping as needed.

Patients without space-occupying lesions fell into two categories. Those with phase I and II evidence for possible medial temporal localization followed a specific protocol detailed later. Those without space-occupying lesions with suspected neocortical epilepsy always underwent phase III intracranial EEG for localization, after phases I and II and mapping (extraoperatively as part of phase III) as needed, if localization provided the possibility of a resection in proximity to functional cortical regions.

In the group of patients considered possible medial temporal by the noninvasive studies, surgery for resection of one anteromedial temporal region would proceed without intracranial study if patients met three major (a–c) or two major and two minor (d–h) criteria from the following list, which were concordant to a single temporal lobe, and no contradictory localization was obtained in any major or minor criterion.

Medial temporal lobe epilepsy: In the group with evidence of a medial temporal onset, resective surgery could proceed without phase III investigations if patients met the following major (a–c) and, as applicable, minor (d–h) criteria: (a) lateralized interictal epileptiform discharges (IEDs) had to be maximal at Sp1, T1, F7, or F9 or the homologous electrodes contralaterally. If IEDs were apparent elsewhere, then we required that ≥90% of all IEDs observed be in the field described earlier. This assessment was based on a sample of ≥50 IEDs; (b) ictal EEG recordings localized the onset of seizures unilaterally to the temporal lobe; (c) unilateral MTS was demonstrated on MRI (hippocampal atrophy with increased signal on T2 images); (d) memory asymmetry on intracarotid amobarbital testing based on defined limits for the number of stimuli recalled depending on the particular protocol and number of stimuli administered; (e) temporal hypoperfusion with interictal SPECT if performed; (f) temporal hypometabolism demonstrated on interictal FDG-PET if performed; (g) neuropsychological profile consistent with hemisphere-specific mesial temporal lobe deficit based on standard criteria; and (h) polymorphic focal slowing confined to the anterior temporal region during ≥50% of the waking interictal record.

If these criteria were not met, then intracranial recording (phase III) was considered necessary.

Although each center varied somewhat in its application of the guidelines in this protocol, an intercenter agreement analysis demonstrated that the ultimate decision, whether to do surgery and what type of resection to do, was excellent (9). The findings of that analysis emphasized the feasibility of performing a multicenter study with a complex evaluation protocol.

After the evaluation was complete, patients were determined to be eligible for resective surgery or not. Those who were offered surgery could choose whether to have surgery. Patients who were determined to be ineligible for resective surgery were sometimes offered callosotomy or multiple subpial transections without resection. These are not considered “resective surgical” patients in this study and are counted in the nonsurgical group. Another group of patients chose not to have surgery after having gone through a complete evaluation, or they terminated the evaluation part-way through the process.

Statistical analyses

Simple bivariate analyses were performed to identify differences between patients who did and did not have surgery. The χ2 test was used for dichotomous or unordered multichotomous variables (e.g., race, surgical site). For ordered grouped variables (e.g., age group), the χ2 test for trend was used. t-Tests were used to test mean differences between surgical and nonsurgical groups for continuous variables (e.g., IQ scores).


A total of 788 patients were recruited to be in the study during the enrollment period. Of these, 223 (28%) were excluded from full consideration for a variety of reasons: six (<1%) died; 54 (7%) did not, on further evaluation, meet basic study eligibility requirements; 34 (4%) withdrew or were lost to follow-up; 50 (6%) patients had incomplete or absent presurgical information; and a decision was made to exclude data from 79 (10%) cases because of inconsistent data collection and storage per study protocols. All but five of these were enrolled early in the study at one site by a single research associate. These represented consecutive patients at that site and consequently were most likely representative of the 565 included for description in this report.

Of the 565 eligible patients for whom essentially complete baseline information was available, 396 (70%) had resective surgery, and 169 (30%) did not. The median time from initiating the evaluation to surgery was 4 months (interquartile range, 2–8 months). Forty-one (10%) of surgical patients had surgery more than a year after initiating their surgical evaluations. Of the169 patients who did not have surgery, 88 were determined not to be appropriate candidates for resective surgery, because either they were not sufficiently localizable or the focus for their seizures was in an inoperable region. Of these 88, 12 had either a callosotomy or multiple subpial transaction procedure. The remaining 81 patients withdrew from further evaluation. Information was not available about what led the patient to terminate the presurgical evaluation (including the counseling about the likely success of the surgery). Ultimately the withdrawn group was similar in many respects to the fully evaluated but ineligible for surgery group, so we have combined all nonresective patients into a single nonsurgical group.

Basic demographic characteristics of the 565 patients in the final cohort are presented in Table 1 along with the number and percentage that had surgery. Few differences were observed between those who did and did not have surgery. A small effect of race/ethnicity was suggested, in that Hispanic patients (52%) were the least likely to go to surgery, and African Americans (63%) somewhat less likely to have surgery compared with white (73%) and Asian (74%) patients (overall p value = 0.03). In addition, married patients were slightly more likely than unmarried patients to have surgery (75 vs. 67%; p = 0.04). A comparison between sites (not shown) demonstrated expected differences in ethnic distributions given the geographic locations of and communities served by the different sites. Some minor differences with respect to education and employment were also observed across sites. With respect to age at onset, age at evaluation, and gender, the patients seen at the centers were quite comparable.

Table 1. Demographic characteristics of patients included in the study overall and by whether resective surgery was performed
Demographic characteristicsN (%) of cohort (n = 565)Resective surgery (n = 396)Nonsurgical (n = 169)p Value
  1. aMissing information for age at onset (n = 25), race (n = 5), education (n = 4), employment status (n = 5), marital status (n = 5).

  2. bThe p value is for a χ2 test for trend.

 Male271 (48)191 (70)80 (30) 0.84 
 Female294 (52)205 (70)89 (30) 
Age at onset (yr)a 
 <5110 (20) 87 (79)23 (21) 0.04b
  5–9 92 (17) 68 (74)24 (26) 
 10–19186 (34)129 (69)57 (31) 
 20–29 83 (15) 54 (65)29 (35) 
 30–3947 (9) 31 (66)16 (34) 
 40+22 (4) 16 (73) 6 (27) 
Age at study entry (yr) 
 <2043 (8) 28 (65)15 (35) 0.38b
 20–29118 (21) 77 (65)41 (35) 
 30–39189 (33)137 (72)52 (28) 
 40–49140 (25)104 (74)36 (26) 
 50+ 75 (13) 50 (67)25 (33) 
 White440 (78)321 (73)119 (27)  0.03 
 African American 57 (10) 36 (63)21 (37) 
 Hispanic40 (7) 21 (52)19 (48) 
 Asian/Native American/other23 (4) 17 (74) 6 (26) 
Highest level of educationa 
 <High school 86 (15) 57 (66)29 (34) 0.11 
 High-school graduate350 (62)242 (69)108 (31)  
 College graduate (4-year degree)123 (22) 96 (78)27 (22) 
 Other  2 (<1)0    2 (100) 
Employment statusa
 Full-time student42 (8) 27 (64)15 (36) 0.81 
 Employed full- or part-time243 (44)175 (72)68 (28) 
 Unemployed218 (39)150 (69)68 (31) 
 Homemaker28 (5) 22 (79) 6 (21) 
 Retired10 (2)  7 (70) 3 (30) 
 Other19 (4) 13 (68) 6 (32) 
Marital statusa 
 Single330 (59)221 (67)109 (33)  0.04 
 Currently married/cohabitating230 (41)173 (75)57 (25) 
 Columbia (New York, NY) 66 (12) 50 (76)16 (24)<0.001
 Montefiore (Bronx, NY)33 (6) 16 (48)17 (52) 
 Jefferson (Philadelphia, PA)134 (24) 86 (64)48 (36) 
 Rochester (Rochester, NY) 92 (16) 81 (88)11 (12) 
 Yale (New Haven CT)106 (19) 64 (60)42 (40) 
 NYU (New York, NY)104 (18) 71 (68)33 (32) 
 MINCEP (Minneapolis, MN)30 (5) 28 (93)2 (7) 

Some of the more salient clinical characteristics in this cohort and their association with surgical decision are presented in Tables 2 and 3. Left-handed patients were less likely to have surgery, as were patients with normal MRIs. Of the 228 with unilateral hippocampal atrophy, 195 (86%) had surgery, which was slightly more than the 50 (75%) of 67 of those with other forms of focal pathology (p = 0.04). Of those with diffuse, bilateral, or multifocal imaging findings, 12 (57%) of 21 with bilateral hippocampal atrophy and nine (64%) of 14 with other nonfocal pathology had surgery (p = 0.67).

Table 2. Clinical characteristics of the cohort overall and by whether surgery was performed
Clinical characteristicsN (% of total)Resective (n = 396)Nonsurgical (n = 169)p Value
  1. aMissing information on handedness (n = 2); duration of epilepsy before surgical evaluation (n = 3); duration of intractable epilepsy (n = 95); specific seizure types (n = 3); seizure frequency (n = 2), history of febrile seizures (n = 2); hippocampal atrophy (n = 79 because scan was either unavailable or of too poor quality), video-EEG (n = 14), etiology (n = 1), intracranial monitoring (n = 7).

  2. bThe p value is for a χ2 test for trend.

 Right464 (82)341 (73)123 (27) 0.001 
 Left82 (15)44 (54)38 (46) 
 Ambidextrous/undetermined17 (3) 11 (65)6 (35) 
Duration of epilepsy (yr) before evaluation for surgerya
 <537 (7) 24 (65)13 (35)0.03b
 5–974 (13)44 (59)30 (41) 
 10–1474 (13)55 (74)19 (26) 
 15–1992 (16)59 (64)33 (36) 
 20+285 (51)213 (75)72 (25) 
Duration of intractable epilepsy (yr since second drug failure)a
 <5140 (30)100 (71)40 (29)0.50b
 5–980 (17)53 (66)27 (34) 
 10–1480 (17)62 (78)18 (22) 
 15–1976 (16)52 (68)24 (32) 
 20+94 (20)71 (76)23 (24) 
History of seizure types (not mutually exclusive)a
Generalized tonic–clonic
 None140 (25)94 (67)46 (33) 0.34  
 Ever had422 (75)301 (71)121 (29) 
Complex partial
 None38 (7) 19 (50)19 (50) 0.005 
 Ever had524 (93)376 (72)148 (28) 
Simple partial
 None455 (81)321 (71)134 (29) 0.77  
 Ever had107 (19)74 (69)33 (31) 
Multiple seizure types
 One seizure type only140 (25)91 (65)49 (35) 0.11  
 >1 seizure type422 (75)304 (72)118 (28) 
Isolated auras
 None317 (56)212 (67)105 (33) 0.04  
 Ever had245 (44)183 (75)62 (25) 
Seizure frequency per mo during 3 mo before entry (does not include isolated auras)a
 1/mo45 (8) 25 (56)20 (44) 0.04  
 2–4/mo166 (29)114 (69)52 (31) 
 5–30/mo267 (47)194 (73)73 (27) 
 ≥30/mo85 (15)62 (73)23 (27) 
Febrile seizuresa
 No427 (76)289 (68)138 (32) 0.03  
 Yes135 (24)105 (78)30 (22) 
Diagnostic imaging
Overall MRI findings
 Normal130 (23)58 (45)72 (55)<0.0001
 Abnormal342 (61)276 (81)66 (19) 
 Equivocal28 (5) 20 (71)8 (29) 
 Inadequate/missing MRI65 (11)42 (65)23 (35) 
If abnormal
 Focal307 (90)255 (83)52 (17) 0.001 
 Diffuse/Multifocal/Bilateral35 (10)21 (60)14 (40) 
Hippocampal atrophy (assessable in 486 scans)
 Absent237 (49)138 (58)99 (42)<0.0001
 Unilateral228 (47)195 (86)33 (14) 
 Bilateral21 (4) 12 (57)9 (43) 
In hospital video-scalp EEG monitoringa
 Unlocalized183 (33)92 (50)91 (50)<0.0001
 Localized368 (67)296 (80)72 (20) 
If localized
 Regional183 (50)134 (73)49 (27) 0.0005
 Focal185 (50)162 (88)23 (12) 
 Variable102 (28)71 (70)31 (30) 0.001 
 Consistent266 (72)225 (85)41 (15) 
 Identifiable etiology199 (35)149 (75)50 (25) 0.07  
 No identifiable etiologic factor (cryptogenic)365 (65)247 (68)118 (32) 
 Cryptogenic, −HAa159 (49)84 (53)75 (47)<0.0001
 Cryptogenic + HAa165 (51)140 (85)25 (15) 
Intracranial monitoringa
 Not done369 (66)233 (63)136 (37)<0.0001
 Performed189 (34)161 (85)28 (15) 
Table 3. Mean quality of life, depression, anxiety, and IQ scores as measured during the presurgical evaluation period
Scale/instrument (no. with valid responses)SurgicalNonsurgicalp Values
  1. aScales range from 0 to 65. Higher scores reflect worse depression or anxiety.

  2. bReported here as T-scores relative to sample of the general U.S. population having a mean of 50 and a standard deviation of 10. Higher scores reflect better health-related quality of life.

  3. cReported here as T-scores relative to a reference sample having a mean of 50 and standard deviation of 10. Higher scores reflect better health-related quality of life.

FSIQ (532)91.488.70.03
PIQ (531)92.489.50.03
VIQ (528)91.889.40.07
Beck's Anxietya (528)10.511.40.41
Beck's Depressiona (504) 9.810.40.47
SF-36–physical healthb (509)46.944.90.03
SF-36–mental healthb (509)44.345.80.16
QOLIE-89–Overallc (508)45.845.70.86
QOLIE-89–Mental healthc (508)47.548.00.62
QOLIE-89–Epilepsy targetedc (508)
QOLIE-89–Cognitive distressc (508)47.747.80.90
QOLIE-89–Physical healthc (508)46.745.80.41

The overall results of inpatient video scalp EEG monitoring were also strongly associated with the surgical decision. Patients with localizable abnormalities were far more likely to have surgery than were those without (80 vs. 50%; p < 0.0001). If the overall scalp EEG findings were localized, patients whose scalp EEG abnormalities were focal versus regional and consistent versus variable localization were more likely to have surgery. The group with the highest proportion going to surgery had clearly focal MRI findings and well-localized scalp EEG abnormalities [116 (92%) of 126].

Other factors associated with a higher probability of having surgery included a history of complex partial seizures, isolated auras, relatively higher recent seizure frequency, and a history of febrile seizures, all probably markers for mesial temporal lobe epilepsy. A small difference in IQ scores was evident, with surgical patients having slightly higher average scores than nonsurgical patients (Table 3).

We examined the drug treatments used in this cohort before study entry, changes in treatment during the evaluation period, and treatment at discharge after surgery (or decision not to perform surgery). The total number of drugs ever tried ranged from two to 14, with a median of five. Somewhat more than half of the cohort (54%) had previous trials of four or all five of these standard therapies: CBZ (95%), PHT (92%), VPA (74%), PB (65%), and PRM (28%). In addition to the older drugs, 86% of the study group already had used one or more of the recently approved or readily available AEDs: lamotrigine (LTG; 51%) and gabapentin (GBP; 64%) being the most commonly used. This probably reflects that they were among the first of the more recent drugs to be approved. Felbamate (FBM) and topiramate (TPM; both 28%) were the next most frequent. The remaining drugs were used in ≤10% of the cohort before initiating an evaluation for surgery. Finally, 26.5% had trials of various drugs that would not be considered “standard” for partial epilepsy [such ethosuximide (ESM), ethytoin (ETH), or benzodiazepines (BZDs)] or experimental drugs [such as pregabalin (PGB) or progabide].

At the time of study entry, 23% of patients were receiving monotherapy; 53%, bitherapy; and 24% were taking three or more AEDs. Ninety percent were taking at least one of the older standard AEDs: CBZ (53%), PHT (29%), VPA (22%), PB (10%), and PRM (5%). Sixty-five percent were also taking one or more of the new drugs. LTG and GBP (both 27%) were the most common. TPM was being taken by 13%, and each of the other drugs was being taken by <5% of the cohort.

During the presurgical evaluation period, 49% of the patients had trials of one or more additional drugs. In 16%, trials were of one or more of the old standard drugs, with CBZ (7%) being the most common. One or more of the newer drugs was tried in 38% of the cohort. The most commonly used newer drugs were LTG (15%), TPM (10%), and GBP and levetiracetam (LEV; each 6%). Other drugs were each used in <5% of the cohort.

Some evidence was seen that use of some of these newer drugs resulted in a slight (∼2 month) delay to surgery. It is not clear that the use of any of these drugs actually resulted in seizure control sufficient that surgery was no longer deemed desirable or necessary.

Although no differences were found in use of a “newer” AED before surgical evaluation over time, the proportion of patients who were taking established drugs at the time of study entry decreased from 94 to 80% (p < 0.001), and the proportion who tried a new AED during the presurgical evaluation period clearly increased from 26% in 1996 to 54% in 2000 (p < 0.0001). Similar trends were observed for drugs being taken after surgery or discharge from evaluation (Table 4).

Table 4. AED use before and at enrollment, during evaluation, and at discharge as well as the proportion going to surgery over the enrollment period 1996–2000
 Year of enrollment in studyp Value trend
1996 (n = 66a)1997 (n = 132)1998 (n = 170)1999 (n = 94)2000 (n = 100)
  1. AED, antiepileptic drug; new AED, drugs approved or readily available in the United States after 1993.

  2. aDrug data missing on three patients for drug history before discharge and for 25 after surgery/end of evaluation.

  3. bAt discharge after surgery or at time a nonsurgical decision was reached for nonsurgical patients.

New AEDs before enrollment60 (91)115 (87)147 (86)77 (82)86 (86)NS
Taking an established AED at enrollment62 (94)122 (92)157 (92)83 (88)80 (80) 0.001 
Taking a new AED at enrollment41 (62) 86 (65)112 (66)61 (65)65 (65)NS
Established AED added during evaluation period11 (17) 20 (15) 23 (14)17 (18)20 (20)NS
New AED added during evaluation period17 (26) 36 (27) 69 (41)39 (41)54 (54)<0.0001
Established AED at dischargeb60 (95)111 (87)144 (89)70 (77)65 (68)<0.0001
New AED at dischargeb35 (56) 78 (61)117 (72)67 (74)79 (82)<0.0001
Had surgery47 (71) 93 (69)126 (74)67 (71)63 (63)NS

The availability and use of the newer AEDs did not seem to affect the proportion of patients who went on to have surgery, and this proportion remained relatively constant over the 5-year recruitment period.


The indications and evaluation for epilepsy surgery have changed significantly over the last 10–15 years. This makes it all the more difficult to align the scattered pieces of information available in the literature, as there is little indication of the overall selection and evaluation process in most reports. We have provided a detailed description of a large cohort of patients from seven established surgical centers who initiated and largely completed evaluations for resective epilepsy surgery. Our data describe the distribution of salient factors that influenced selection for surgery and therefore served as filters in the selection process. Many of these factors may not come as a surprise (presence of a focal abnormality on the MRI), but it is nonetheless helpful to quantify the number of patients who do and do not have various clinical or demographic factors to provide a better understanding of how many and which patients with intractable focal epilepsy are considered appropriate for surgery under current evaluation protocols. Such information also provides an estimate of the number of people with intractable partial epilepsy who are currently not good surgical candidates and who are therefore in need of other forms of treatment, advanced forms of evaluation, and new surgical approaches.

The clinical factors associated with having surgery included right-handedness, a history of febrile seizures, focal findings on MRI, and consistent, focal scalp EEG abnormalities. Most of these factors reflected unilateral hippocampal atrophy. Of note, a large proportion of patients (often ≥50%) without well-localized scalp EEGs or MRIs still had surgery. In those who had invasive monitoring, even in the absence of a clear localizing MRI abnormality, 70–80% had surgery. Careful follow-up of this cohort is ongoing and will help determine whether the seizure outcomes are compromised in those for whom the initial evidence was not clearly localizing.

IQ scores were only slightly lower in the nonsurgical group. This, along with being left-handed or ambidextrous, may reflect patients with more diffuse brain disease (10). Conversely, quality-of-life scores, although overall a half standard deviation less than expected (based on comparisons to a reference sample of individuals with epilepsy who have a broader range of disease severity) (5,11), were not associated with the decision to have surgery. The same was true for depression and anxiety scores. This suggests that these were neither an impediment to surgery (anxiety and depression in particular) nor, within the context of a group that had already self-selected to have a surgical evaluation, did they influence the surgical decision (quality of life in particular).

Although gender and age at study entry did not influence surgical decision, age at onset had a small effect, largely due to those with childhood onset being slightly more likely to have surgery. This is primarily due to temporal lobe epilepsy associated with hippocampal atrophy, whose onset is more likely to occur early rather than later in life (12). Of possible concern is the finding that African American and especially Hispanic patients are less likely to have surgery than are white and Asian (or other) individuals. To the extent that we were able to examine this issue, these apparent racial differences seemed to persist at each site and raise some concerns regarding inequalities in access to or preferences for medical care. Other reports in the literature indicate racial differences in the use of other surgical procedures, although in at least one carefully conducted and analyzed study, these differences ultimately were due to underlying clinical differences across racial and ethnic groups and not behavioral or social factors (13). Nonetheless, this issue deserves close attention and further study.

The nonsurgical group contained patients who were found to be poor surgical candidates as well as those who terminated the evaluation procedures early before appropriateness for surgery had been fully determined. We do not know the extent to which this decision was influenced by counseling about the likelihood that surgery would occur, had these patients proceeded with the full evaluation. Whereas the early termination group may contain some patients who ultimately would have been considered good surgical candidates had they been thoroughly evaluated, their clinical profiles with respect to factors such as focal findings and hippocampal atrophy in particular on the MRI were much closer to the group fully evaluated and deemed inappropriate for resective surgery than they were to the group ultimately offered surgery.

All sites agreed to follow a similar protocol for evaluating patients. It was impossible to standardized practice completely across the seven sites, an intercenter agreement study indicated that, for a given patient, there was excellent agreement on the ultimate decision of whether to offer surgery and what kind of procedure to use (9). Consequently, the differences in proportions receiving surgery may reflect differences in the types of patients referred to different sites (more difficult cases being referred to the more experienced centers) rather than differences in how the sites treated patients. In addition, there may have been some local variation in the physicians referring to the different sites. As there is no one system in the United States for referring patients for surgery, different patients were first treated and evaluated by physicians with a wide range of expertise who may have varied in the efficiency with which they selected patients for referral for surgery.

We had the unique opportunity of recruiting during a time when several new AEDs were approved and came into widespread use. Although never intended to assess the effectiveness of these new drugs, our observational data provide some documentation of their use and perhaps usefulness in this situation. Over the time we recruited patients into the study (1996–2000), the proportion of patients who went on to have surgery remained relatively constant at ∼70% (Table 4). The small dip in 2000 may simply be due to our closing out the cohort and not accepting any new surgical patients >1 year after their initial enrollment date. Although patients may have experienced some improvement in seizure control with the addition of new AEDs, ultimately no evidence suggested that the newer drugs provided sufficient seizure control to obviate the need for surgery. Such a finding supports the proposition that early rather than later surgical intervention should be considered because continued trials of different AEDs are unlikely to produce satisfactory seizure control. There is the hope that effective early rather than later intervention will allow an individual to resume a relatively normal life. This is part of the premise of the new Early Randomized Surgical Epilepsy Trial (ERSET), which will randomize patients to surgery or continued medical therapy at a relatively early stage in the expression of medically refractory temporal lobe epilepsy (14–16). For patients who are not good candidates for resective surgery, the possibility of other procedures, such as multiple subpial transection, a procedure used in a few of the “nonsurgical” patients in this series, needs further exploration and study. Early results of case series are promising, and a high-quality randomized trial is warranted (17).

Ongoing follow-up of this cohort will permit us to identify predictors of seizure, quality of life, psychiatric, and neuropsychological outcomes after surgery in a large, well-defined cohort of individuals who were evaluated in the 1990s at seven established surgical centers. The description of the baseline characteristics of this cohort and identification of factors that influenced the decision to proceed with surgery provide an essential framework that will allow comparison with other studies both present and future and facilitate the interpretation and generalization of our results to other settings. Although surgical intervention is typically delayed for many years, our findings suggest that further efforts to control seizures with new drugs in patients for whom numerous other drugs have failed, rarely succeed. This tends to give further support to the idea that surgical intervention should be tried earlier in the course of epilepsy once pharmacoresistance has been documented (16).


Acknowledgment:  This study was supported by grant RO1 NS 32375 from the NIH-NINDS.