You are on page 1of 9

ARTICULO DE LOBULO TEMPORAL

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x

CRITICAL REVIEW AND INVITED COMMENTARY

Long-term seizure outcome of surgery versus no


surgery for drug-resistant partial epilepsy: A review of
controlled studies
*Dieter Schmidt and yzKnut Stavem

*Epilepsy Research Group Berlin, Germany; yHØKH, Research Centre, Akershus University Hospital,
Lørenskog, Norway; and zFaculty of Medicine, University of Oslo, Oslo, Norway

pared to 139 of 1113 (12%) of nonoperated con-


SUMMARY
trols [pooled random effects relative risk (RR)
A majority of patients with formerly drug-resis- 4.26, 95% confidence interval (CI) 3.03–5.98]. The
tant temporal lobe epilepsy become seizure-free pooled risk difference in favor of surgery was 42%
after surgery. However, apart from one 12-month (95% CI 32–51%). We found no comparative out-
randomized trial, it is unclear how many become come data in patients with extratemporal lobe
seizure-free because of surgery. To determine the epilepsy only. The available evidence from mostly
net benefit of surgery, we performed a systematic nonrandomized observational studies indicates
review and meta-analysis of the published evi- that in appropriately selected patients with drug-
dence of how many patients in similar studies resistant temporal lobe epilepsy, the combination
become seizure-free without surgery. Of 155 of surgery with medical treatment is 4 times as
potentially eligible articles reviewed in full text, 29 likely as medical treatment alone to achieve free-
(19%) fulfilled eligibility criteria. After excluding 9 dom from seizures.
publications, 20 studies form the base of evidence. KEY WORDS: Epilepsy surgery, Long-term
Overall, 719 of 1,621 (44%) of patients with mostly outcome, Chronic epilepsy, Antiepileptic drugs,
temporal lobe surgery were seizure-free com- Seizure prognosis.

Adjunctive treatment with resective surgery and a randomized trial, the long-term seizure outcome of
change of medical regimen are both standards of care for surgery versus medical treatment in nonoperated patients
eligible patients with chronic drug-resistant epilepsy (En- is not well known (Wiebe et al., 2001; T!llez-Zenteno
gel et al., 2003; French et al., 2004). A meta-analysis of et al., 2005).
surgical outcome indicates that 2 years after surgery 55% Although most studies have focused on seizure outcome
of patients with formerly drug-resistant temporal lobe in surgical cases, a number of epilepsy centers have
epilepsy are completely seizure free and 68% are free of reported clinical observations of seizure outcomes in sur-
complex partial seizures (Chapell et al., 2003). It is gical versus nonoperated patients, including patients who
unclear, however, how many patients become seizure-free were found unsuitable for surgery (T!llez-Zenteno et al.,
because of surgery. To determine the net-benefit of 2005). The strength of inference that can be derived from
surgery, one would need to know how many patients in these nonrandomized studies is clearly less robust than
similar studies become seizure-free without surgery that from the only randomized trial; however, data from
(Chapell et al., 2003). However, apart from one 12-month cohort studies still inform about clinical practice in many
areas, including epilepsy. Meta-analyses of observational
Accepted October 29, 2008; Early View publication February 23, data have a role in medical effectiveness research, pro-
2009. vided that the possible sources of heterogeneity between
Address correspondence to Prof. Dr. Dieter Schmidt, Epilepsy
Research Group, Goethestr.5, D-14163 Berlin, Germany. E-mail: studies and the influence of biases or confounding factors
dbschmidt@t-online.de are carefully examined (Egger et al., 1998). Furthermore,
Wiley Periodicals, Inc. the question of long-term outcome may not be amenable
ª 2009 International League Against Epilepsy to randomized trials because of feasibility and ethical

1301
1302
D. Schmidt and K. Stavem

issues. In addition, a comparison of surgical outcome estimates for the studies, presenting the RR and its 95%
versus a change in medical regimen has become clinically CI, with values >1 favoring surgery. We aggregated the
important, because recent outcome data in apparently RRs in tables and a forest plot. The analysis was stratified
drug-resistant epilepsy, including in nonsurgical candi- different ways according to subgroups as a sensitivity
dates, suggest that up to 21% of patients may be rendered analysis, because of heterogeneity between studies. We
seizure-free for more than 2 years by a change in regimen had considered assessing the influence of different vari-
(Selwa et al., 2003). Given that epilepsy is drug-resistant ables in a meta-regression analysis, but we realized that
in as many as one in three newly diagnosed patients the limited number of studies, the heterogeneity between
(Kwan & Brodie, 2000; Schmidt & Lçscher, 2005), uncer- the studies, and association between various variables
tainty about the comparative efficacy of surgery versus a would raise a number of problems that could not be
change of medical regimen in nonoperated patients, the resolved. Therefore, we have abstained from presenting a
two commonly used treatments for chronic epilepsy, is a meta-regression and instead presented RRs for a number
strategic problem. This scenario prompted us to examine of subgroups. Two studies reported more than one follow-
the evidence for long-term seizure freedom following sur- up period (McLachlan et al., 1997; Picot et al., 2008), and
gery and after drug treatment in nonoperated patients. one had separate reports for different time intervals (Sta-
vem & Guldvog, 2005; Stavem et al., 2008). For these
studies, we assessed each follow-up period separately.
Methods Hence, in one comparison, we included studies with sev-
Data sources eral follow-up periods in the same population, presenting
A medical librarian performed a comprehensive litera- results for £24 months and >24 months. In the remaining
ture search of the PubMed, Medline, Embase, Index analyses, we included only results from the longest fol-
Medicus, and Cochrane databases (search strategy in low-up time for each study, presenting results according
Appendix). Literature searches were restricted to full- to length of follow-up (£24 months and >24 months),
length articles published between January 1947 and June publications with and without zero seizure-free outcome
2007 in English, German, or French. In addition, we also in the medical control group, year of publication (1947–
hand-searched bibliographies of reviews, original articles, 1999 and 2000–2007), surgical topography: temporal lobe
and book chapters, and consulted experts about other versus temporal lobe plus extratemporal lobe, and study
studies. design. Because of lack of information, we could not per-
form a subgroup analysis for duration of treatment prior to
Study selection and classification surgery; definition of seizure-freedom; age at surgery; eti-
Two reviewers independently applied the following ology, mesial temporal sclerosis versus other etiology;
study inclusion criteria: reports of ‡20 patients of any age and choice of nonoperated group, patients on waiting lists
undergoing resective epilepsy surgery utilizing a control for surgery versus those found ineligible for surgery. We
group, outcomes reported after a mean/median follow-up also summarized the RRs of being without AEDs after
of ‡1 year after surgery or initial evaluation, quantitative surgery compared to no surgery at the longest follow-up
report of seizure freedom with or without concurrent in the studies (n = 4) that reported this, using a similar
antiepileptic drug (AED) treatment, and description of type meta-analytic technique.
of surgery and number of patients undergoing each inter- The meta-analyses were done using random effects
vention. We further classified studies by surgical topogra- models, aggregating studies with a weighting equal to the
phy into those studies reporting outcomes of temporal lobe inverse of the variance of the estimate for the study. Ran-
surgery, and those grouping temporal and extratemporal dom effects models were chosen because of heterogeneity
surgery together. We found a total of 31 publications and between studies, although fixed effects models would
excluded 11 articles. The remaining 20 publications form have yielded essentially similar results. The weights are
the base of evidence for this review (see results). not shown on the forest plots or in the tables. We assessed
interstudy variability (heterogeneity) by describing I2, that
Data gathering is, the percentage of total variation across studies that is
Two reviewers independently abstracted all data, caused by heterogeneity rather than chance (Higgins et al.,
resolving disagreements through discussion. We accepted 2003). A percentage of 25%, 50%, and 75% has been sug-
outcome definition as used by authors in each study. These gested to indicate low, medium, and high heterogeneity
usually referred to seizure outcome in terms of seizure (Higgins et al., 2003). In addition we present the results of
freedom. a test for heterogeneity using DerSimonian and Laird’s Q
(DerSimonian & Laird, 1986). Because of heterogeneity
Data analysis between studies, we present results for subgroups rather
For comparison of seizure-free outcome between than for the aggregate of all studies. For seizure outcome
surgical versus nonsurgical groups, we produced summary analysis of surgery and medical treatment, we
Epilepsia, 50(6):1301–1309, 2009
doi: 10.1111/j.1528-1167.2008.01997.x
1303
Epilepsy Surgery versus No Surgery

used seizure-free versus not seizure-free, based on the Study population and trial design
classifications used in the included papers. The authors of We identified 20 controlled studies (Table 1).
the papers used Engel’s outcome classification (Engel Except for one (Wiebe et al., 2001), all studies were
et al., 2003), modifications of this, or their own outcome nonrandomized (Table 1). Two studies reported more
classifications. The overall results are also presented using than one follow-up period (McLachlan et al., 1997;
a similar metaanalysis with risk differences as outcome. Picot et al., 2008), and one had separate reports for
We used the Metan procedure in Stata version 10 for all different time intervals (Stavem & Guldvog, 2005;
analyses (Stata Corp., College Station, TX, U.S.A.). Stavem et al., 2008). The outcome of temporal lobe
surgery only was assessed in 17 of 20 studies, the
other three studies reported combined outcome of
Results temporal and extratemporal lobe epilepsy (Table 1).
Evidence base We found no studies on seizure outcome for
The literature search yielded 155 references, of which extratemporal lobe surgery only versus AED treatment
31 (20%) were potentially eligible and were reviewed without surgery.
independently by two reviewers. Twenty (13%) fulfilled In 16 of the 20 studies, controls were patients evaluated
the eligibility criteria and constitute this analysis’ data set for surgery but who did not undergo resective surgery,
(Fig. 1). three studies did not describe surgical eligibility of their
control group, and in one study patients were randomized
as controls before surgical eligibility was assessed
(Table 2). Participants in the medical control group who
were evaluated for surgery either failed to meet criteria for
surgery, declined surgery, or were waiting for surgery and
continued to receive AED treatment (Table 2). In three
early studies, controls were patients treated with AEDs
undergoing craniotomy without excision of brain tissue
(Penfield & Steelman, 1947; Penfield & Paine, 1955;
Ommaya, 1963).
Definitions of seizure-freedom in the surgical studies
fell into four main categories. In most studies (8 of 20
studies, 40%) the measure for seizure freedom was unde-
fined. In 5 of 20 studies (25%), Engel’s classification was
used (Engel et al., 1993), 1 study used in addition the
ILAE classification (Wieser et al., 2001); in 3 of 20 stud-
ies, patients were free of any seizures and auras, whereas
in a further four studies, patients having auras were con-
sidered to be seizure-free (Table 1).
The definition of seizure-freedom for surgical and
Figure 1. nonsurgical treatment outcome is given in Table 1.
Search strategy and evidence base. *Reasons for Ninety percent of the studies focused on seizure free-
exclusion were: duplicate publications, that is, studies dom at the last follow-up or the last year, 5% did
sharing the same or overlapping patient populations at not specify a time frame, and 5% offered no defini-
a similar follow-up period (Bien et al., 2001; tion. In 90% of the studies, seizure-free outcome was
Helmstaedter et al., 2003, overlapping with Bien et al., determined after the first surgery; in two studies 5%
2006; Haglung & Moretti Ojemann, 1993, overlapping of patients in each study underwent a second surgical
with Moretti Ojemann & Jung, 2006; Guldvog et al., procedure (Vickrey et al., 1995; Bien et al., 2006). In
1994; Stavem & Guldvog, 2005, overlapping with 20% of studies, a change in AEDs was reported after
Stavem et al., 2008; Mikati et al., 2004; had overlapping surgery (Penfield & Steelman, 1947; Vickrey et al.,
populations with Mikati et al., 2006; Picot et al., 2004 1995; Wiebe et al., 2001; Yasuda et al., 2006).
had overlapping populations with Picot et al., 2008; no Although the total number of AEDs was lower in the
quantitative seizure outcome data for the nonoperated surgical group, a dose increase of AEDs was reported
group F: Rausch & Crandall, 1982; Baxendale & in two studies (Wiebe et al., 2001; Yasuda et al.,
Thompson, 1996; Kellett et al., 1997; follow-up of less 2006). In one study, surgically treated patients who
than 1 year E: (Aydemir et al., 2004). became seizure-free after surgery after a change of
Epilepsia ILAE medical regimen, including a modern AED, were
excluded (Kumlien et al., 2002).
Epilepsia, 50(6):1301–1309, 2009
doi: 10.1111/j.1528-1167.2008.01997.x
1304
D. Schmidt and K. Stavem

Table 1. Study population and trial design in 20 controlled studies comparing seizure-free
outcome in operated patients and nonoperated controls. The number of surgical patients and
controls is given in Figure 2
Seizure-free Seizure-free
outcome outcome
Study measures Type of Follow-up measures Seizure-free
Authors (year) design (surgical) surgery (years) (nonsurgical) (nonsurgical)
McLachlan et al. (1997) C/CR No auras TL 1, 2 Undefined No
Gilliam et al. (1999) C/CR No auras, with auras TL 2 No auras, with auras No
O’Donoghue et al. (1998) C/CR Undefined TL 2 Undefined Yes
Altshuler et al. (1999) C With auras TL 4–13 With auras Yes
Ommaya (1963) C/CR No aura TL 4–13 No auras No
Penfield and Steelman (1947) C Undefined TL 4–13 Undefined No
Penfield and Paine (1955) C/CR Undefined TL 4–13 Undefined No
Vickrey et al. (1995) C/CR No auras, with auras TL + XTL 4–13 No auras, with auras Yes
Markand et al. (2000) C Engel class I, no auras TL 1 No auras No
Wiebe et al. (2001) RCT No auras, with auras TL 1 No auras, with auras Yes
Yasuda et al. (2006)a C Engel class IA TL 1 Engel class IA Yes
Mikati et al. (2006) C Engel class IA TL 2 Engel class IA Yes
Bien et al. (2006) C/CR Undefined TL 4–13 Undefined Yes
Derry et al. (2001) C Undefined TL 4–13 Undefined Yes
Jones et al. (2002) C No auras TL 4–13 No auras Yes
Kumlien et al. (2002) C Undefined TL 4–13 Undefined Yes
Li (2002) C/CR Undefined TL 4–13 Undefined No
Moretti Ojemann and Jung (2006) C Author specific TL 4–13 with auras Yes
Stavem et al. (2008) C/CR Undefined TL + XTL 4–13 Undefined Yes
Picot et al. (2008) C Engel I, ILAE TL + XTL 1, 2, 4–13 Engel I, ILAE Yes
a
Additional data supplied by study author on request.
C, cohort; CR, cross-sectional; ILAE, ILAE, International League Against Epilepsy; RCT, randomized controlled trial; TL, outcome
for temporal lobe surgery only; TL + XTL, combined outcome after temporal and extratemporal lobe surgery.

Seizure-free outcome in studies comparing surgical Factors that may influence seizure-free outcome
versus drug treatment In a sensitivity analysis, we explored the impact of
The weighted pooled proportion of long-term sei- potential explanatory factors for seizure-free outcome
zure-free patients was determined in a total of 20 of surgical versus medical studies in the study design
studies comparing surgical versus medical outcome and the study population (Table 3). The heterogeneity
(Fig. 2). Overall, 719 of 1,621 patients (44%) with could also be reduced in some of the subgroups. For
mostly temporal lobe surgery were seizure free com- assessment of outcome in different study designs we
pared to 139 of 1,113 (12%) of nonoperated controls included: duration of follow-up, random controlled
(RR 4.26, 95% CI 3.03–5.98). The risk difference trial (RCT) versus cohort or cohort/cross-sectional,
between surgery and medical treatment without sur- and the effect of excluding seven medical studies
gery was 42% (95% CI 32–51%) in favor of surgery. with zero seizure-free outcome (Table 3). For assess-
Assessment of heterogeneity by the I2 method showed ment of outcome in a different study population we
moderate heterogeneity of 60% (Fig. 2), and Q was included: temporal lobe surgery versus studies report-
47.05 (p < 0.001). ing a grouped outcome of temporal plus extratempo-
In four studies that reported how many patients ral lobe surgery, studies of surgery published before
were off AEDs at the end of follow-up, more patients 1999 (Table 3). In all subgroups surgery had a statis-
were off AEDs in the surgical group versus the non- tically better outcome than continued medical treat-
surgical group; 32% vs. 6%, with an estimated ment. None of the examined factors seemed to have
pooled RR for being off AEDs with surgery of 4.67 a great influence of the seizure-free outcome of surgi-
(95% CI 2.18–10.01) compared to medical treatment cal versus no surgical treatment (Table 3). We did
only (McLachlan et al., 1997; Altshuler et al., 1999; not find sufficient data to explore other potential
Derry et al., 2001; Stavem et al., 2008). Here, hetero- factors such as age, reasons for patients not to have
geneity between studies was low with I2 of 0% and surgery, specific etiologies, or different definitions of
Q of 0.40 (p = 0.94). seizure-free outcome.

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x
1305
Epilepsy Surgery versus No Surgery

Table 2. Description of medical controls in the included studies. Participants in the medical
control group who were evaluated for surgery either failed to meet criteria for surgery (U),
declined surgery (R), or were waiting for surgery (W) and continued to receive antiepileptic drug
(AED) treatment. In three early studies, controls were patients undergoing craniotomy without
excision of brain tissue (S), and in three studies surgical eligibility of controls was not described (N).
The number of control patients for each study is given in Figure 2
Authors (year) Description of controls
McLachlan et al. (1997) Patients evaluated for surgery, but not operated on for various reasons (R, U)
Gilliam et al. (1999) Patients on waiting list after being evaluated and selected for surgery with anterior temporal lobectomy
in 1995 and 1996 (W)
O’Donoghue et al. (1998) Patients with definite epilepsy seen consecutively in an epilepsy clinic with a follow-up of 1 year. A total of
374 patients from 183 consecutive with definite epilepsy at an epilepsy follow-up clinic with ‡1 year follow-up,
and 191 consecutive patients referred for video-EEG telemetry for assessment of epilepsy surgery
(excl. those with IQ < 70). Surgical ineligibility of controls was not described (N)
Altshuler et al. (1999) Patients undergoing presurgical evaluation, who did not have surgery for lack of localized seizure focus
(12 of 13) or extratemporal focus not suitable for anterior temporal lobe resection (1 of 13) (U)
Ommaya (1963) Patients with a diagnosis of temporal lobe epilepsy presenting for surgery, were evaluated and underwent
craniotomy, but no removal of brain tissue because of ‘‘lacking focal activity’’ (U, S)
Penfield and Steelman (1947) Patients operated on 1939–1944 with craniotomy and cortical excision—treatment group (n = 59) versus
craniotomy without cortical excision (controls) (n = 16) (U, S)
Penfield and Paine (1955) Patients with negative explorations operated on between 1945 and 1950, versus surgery patients (U, S)
Vickrey et al. (1995) Patients were evaluated for surgery, but not operated on; no identified focus in 42 (91%), two contraindications
for surgery, two withdrew from surgery or evaluation (R, U)
Markand et al. (2000) Patients evaluated for surgery who were considered unsuitable for anterior temporal lobe resections or
elected against surgery (R, U)
Wiebe et al. (2001) Patients were randomized to a delay of presurgical evaluation for 1 year or surgery (W)
Yasuda et al. (2006)a Medically treated patients with mesial temporal lobe epilepsy followed for more than 12 months (personal
communication by the study author), waiting for finishing presurgical evaluation, waiting for surgery after
evaluation, or those unsuitable for surgery versus patients undergoing temporal lobe surgery (W, U)
Mikati et al. (2006) Individually matched patients undergoing presurgical evaluation, who were not eligible for surgery (no single
seizure focus, or unacceptable neurologic risk) versus consecutive patients undergoing surgery
(75% temporal lobe resections) (U)
Bien et al. (2006) Patients not eligible for surgery after presurgical assessment (focus not identified, multiple foci, unacceptable
neurologic risk) versus surgical patients (W, U)
Derry et al. (2001) Evaluated but ineligible patients receiving medical management followed for 8.5 years versus surgical patients (U)
Jones et al. (2002) Individuals who were evaluated for surgery but did not proceed to surgery and were receiving AED
treatment versus surgical patients (U)
Kumlien et al. (2002) Medically treated patients with MRI-verified mesial temporal sclerosis (excluding dual pathology, nonepileptic
seizures, nonadherence, uncertain EEG findings) who were considered unsuitable for surgery
(no reasons given) versus surgical patients (U)
Li (2002) Patients evaluated as surgical candidates who declined surgery or were found otherwise unsuitable for
surgery versus surgical patients (U)
Moretti Ojemann and Controls were patients with refractory epilepsy who were given medical treatment. Surgical ineligibility of
Jung (2006) controls not described (N)
Stavem et al. (2008) Patients operated on between 1948 and 1992 with individually matched nonoperated controls hospitalized
for severe epilepsy. Matched for age, sex, and seizure type. Surgical ineligibility of controls not described (N)
Picot et al. (2008) Controls were patients with refractory epilepsy who were given medical treatment. Following presurgical
exploration, 44% of controls were ineligible for surgery (U)
a
Additional data supplied by study author on request.

95% CI 3.03–5.98), or a pooled risk difference of 42%


Discussion (95% CI 32–51%), underscoring that the combination of
The three main results of our review of seizure-free out- surgery with medical treatment is more efficacious than
come in a total of 2,734 patients with drug-resistant medical treatment alone, confirming the result of the only
epilepsy after temporal lobe surgery versus medical treat- randomized study in the literature (Wiebe et al., 2001). In
ment in broadly similar nonoperated patients are first, that addition, in the small number of studies reporting AED-
on average, 719 of 1,621 (44%) of patients with mostly free outcome, more than four times more surgical patients
temporal lobe surgery were seizure free compared to 139 were able to discontinue AEDs compared to nonsurgi-
of 1,113 (12%) of nonoperated controls (pooled RR 4.26, cal controls RR 4.67 (95% CI 2.18–10.01). Second,

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x
1306
D. Schmidt and K. Stavem

Figure 2.
Forest plot of the proportion of patients who are seizure-free after surgical treatment and of nonoperated controls.
Event = seizure-free patients. Seizure freedom is shown as defined by study authors. AEDs may have been
discontinued in either group. The black marker and the horizontal lines show the relative risk (RR) and 95%
confidence interval (CI) for each trial. The size of the shaded square is proportional to the weight of the study (%
weight) in the pooled estimated RR. The diamond represents the pooled RR and 95% CI, using a random effects
model. I2 is a measure of heterogeneity between the studies (see Methods).
Epilepsia ILAE

nevertheless, it is remarkable that seizure-free outcome treatment (Cucherat, 2004; Tellez-Zenteno et al., 2007).
was observed in 12% of similar patients without surgery In one review covering five studies, improvement was
by AED treatment alone. This shows that drug resistance noted in surgical patients compared with controls without
can be overcome by medical treatment in a minority of providing quantitative outcome data (Tellez-Zenteno
patients after having experienced lack of seizure control et al., 2007). An earlier review by Cucherat (2004) cov-
for many years. Third, no evidence is available for seizure- ered four studies, including three studies reviewed by
free outcome of extratemporal surgery versus no surgery. Tellez-Zenteno et al. (2007) and the only randomized
The surgical outcome studies that were included in controlled trial by Wiebe et al. (2001). Cucherat (2004)
our analysis (n = 20) had median 56% (range 7–85%) concluded that the evidence was very limited, except for
seizure-free patients by authors’ definition, which is simi- the trial of Wiebe et al. (2001). The long-term seizure-free
lar to the outcome of the only randomized trial where it outcome of nonsurgical patients with chronic drug-
was reported that among the 40 surgery patients 38% were resistant epilepsy in our review of median 5% (range
completely free of seizures and 58% were free of seizures 0–35%) of the studies (n = 20) is similar to that observed
impairing awareness (Wiebe et al., 2001). We found two for older AEDs in nonsurgical patients in other series,
reviews in the literature covering a small number of which, however, provided no data for surgical patients
studies on seizure outcome of surgical versus medical (14%, Bauer & Burr, 2001; 0%, Huttenlocher & Hapke,

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x
1307
Epilepsy Surgery versus No Surgery

Table 3. Sensitivity analyses with different cross-classifications of studies, showing pooled relative
risks (RRs) of seizure-freedom of surgery versus medical treatment without surgery according to
individual study features. The RRs, which were determined from random effect models are similar
and their 95% confidence intervals (CIs), are overlapping. The table also shows the number of
events (=seizure-free patients) and an index of heterogeneity (see footnote)
No. of Relative risk Events Events
Study feature studies (95% CI) treatment control I2, %d p-Valuee
Follow-up duration, all papers, longest follow-up of
each studya
24 months or less 7 6.90 (2.73, 17.46) 233/421 74/592 80 <0.001
More than 24 months 13 3.82 (2.67, 5.47) 486/1200 65/521 44 0.04
Seizure-free controls onlyb, and longest follow-up 13 3.50 (2.65, 4.63) 494/955 139/916 49 0.02
of each studyb
24 months or less 4 3.44 (2.11, 5.63) 103/203 74/475 47 0.13
More than 24 months 9 3.60 (2.48, 5,22) 391/752 65/441 54 0.03
Follow-up, seizure-free controls only, all papers,
allowing for more than one follow-up period for
each paperc
24 months or less 6 4.09 (2.76, 6.06) 250/431 104/695 60 0.03
More than 24 months 9 3.60 (2.48, 5.22) 391/752 65/441 54 0.03
Year of publication, longest follow-up of each studya
1947–1999 8 4.71 (2.14, 10.38) 332/966 69/572 61 0.01
2000–2008 12 4.39 (2.89, 6.67) 387/655 70/541 62 0.002
Surgical topography, longest follow-up of each studya
Temporal lobe 17 4.77 (3.05, 7.46) 561/1268 107/894 65 <0.001
Temporal and extratemporal lobe combined 3 3.81 (2.31, 6.30) 158/353 32/219 46 0.16
Study design, longest follow-up of each studya
Cohort/randomized controlled trial 11 4.73 (3.05, 7.32) 320/540 49/427 52 0.02
Cohort/cross-section 9 3.71 (2.16, 6.37) 399/1081 90/686 61 0.008
a
Only including the longest follow-up period reported in each study population.
b
Excluding studies with no seizure-free patients in the control group.
c
Two papers presented results for more than one follow-up period.
d
Proportion of the total variance due to between study variance (I2, see Methods).
e
Der Simonian and Laird Q statistic (see Methods).

1990; 10%, Harbord & Manson, 1987). More recently, it nant abnormalities were in the temporal lobe in the medi-
was shown that a significant minority of patients (16.6%) cally treated control subjects and in the surgical group.
were rendered seizure-free by addition of newly adminis- Our data present the best available evidence for long-
tered AEDs, even after failure of two to five past antiepi- term outcome (in the absence of long-term randomized
leptic drugs (Schiller & Najjar, 2008). In a clinical trials), and are informative on the outcome of two stan-
observation of 155 nonsurgical adult patients with chronic dard strategies for treating patients with drug-resistant
epilepsy that had been resistant to previous drug therapy, partial epilepsy. Except for one 12-month randomized
16% of all drug introductions resulted in seizure freedom trial, outcome following surgery versus no surgery was
(defined as seizure freedom at last follow-up for determined in nonrandomized cohort and cohort/cross-
12 months or longer), and 28% of the patients were sectional studies. The cross-sectional design may be a
rendered seizure free by a drug introduction (Luciano & limitation because it does not control for potential con-
Shorvon, 2007). founders associated with surgical intervention (e.g., pla-
Although our large review summarizes the available cebo effect), selection bias, or the possible improvement
literature on seizure-free outcome in a total of more than of a nonsurgical control over time when treated in a com-
2,700 patients with drug-resistant epilepsy undergoing prehensive epilepsy center. Alternative study designs
surgery versus no surgery, it has a number of limitations. using prospective, nonrandomized medical controls or
One of the important limitations of most studies in our presurgical assessments as controls, however, do not
review is the lack of randomization and the inherent bias eliminate selection bias or confounding effects poten-
in the selection of control groups, which because of the tially associated with surgery. An early study, at a time
nature of the surgery may not be entirely comparable to when medical options were limited, attempted to assess
the surgical patients, since there was always a reason that the potential placebo effect of surgery. Although the
they did not proceed to surgery. However, the predomi- inclusion of studies as far back as 1947 has obvious

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x
1308
D. Schmidt and K. Stavem

limitations, the comparison of nonresection surgery and between pooled RR was 42% in favor of surgery. (2)
completeness by showing all available studies speak in Nevertheless, it is remarkable that seizure-free outcome
favor of including early studies. In addition, according to was observed in 12% of nonsurgical patients by AED
the sensitivity analysis, shown in Table 3, publication treatment alone. This suggests that drug resistance can be
data of the study did not appear to influence outcome. overcome by medical treatment in a minority of patients
The Penfield surgical series from 1947 demonstrated that after having experienced lack of seizure control for many
5 of 16 patients (32%) of a ‘‘sham’’ control group who years. (3) An important minority of seizure-free surgical
received an exploratory craniotomy without cerebral patients (but only very few nonsurgical cases) are able to
resection experienced a 50% or higher seizure reduction, discontinue AEDs.
but none became seizure-free (Penfield & Steelman,
1947). The series by Ommaya also explored a nonsurgi-
cal control group of 25 patients undergoing craniotomy
Acknowledgment
without cerebral resection (Ommaya, 1963). Again, none We confirm that we have read the Journal’s position on issues
of the controls became seizure free (Ommaya, 1963). involved in ethical publication and affirm that this report is consistent
with those guidelines.
The series from the Bonn University compared seizure-
free outcome in three nonsurgical control groups and Disclosure: We have no conflict of interest.
found clinically important differences in seizure-free out-
come (Bien et al., 2006). Patients who were unsuitable References
for surgery had the highest seizure-free rate of 24%,
whereas patients awaiting presurgical assessment or Altshuler L, Rausch R, Delrahim S, Kay J, Crandall P. (1999) Temporal
lobe epilepsy, temporal lobectomy, and major depression. J Neuro-
withdrawing from waiting for presurgical assessment had psychiatry Clin Neurosci 11:436–443.
much lower rates of 5% and 14%, respectively (Bien Aydemir N, "zkara C, Canbeyli R, Tekcan A. (2004) Changes in quality
et al., 2006). This study shows that the choice of nonsur- of life and self-perspective related to surgery in patients with tempo-
ral lobe epilepsy. Epilepsy Behav 5:735–742.
gical controls can influence the seizure outcome. In addi- Bauer J, Burr W. (2001) Course of chronic focal epilepsy resistant to
tion, not all patients in the medical and the surgical group anticonvulsant treatment. Seizure 10:239–242.
were treated with AEDs over the years in the epilepsy Baxendale SA, Thompson PJ. (1996) ‘‘If I didn’t have epilepsy…’’
Patient expectations of epilepsy surgery. J Epilepsy 9:274–281.
center doing surgery; many were treated elsewhere Bien CG, Kurthen M, Baron K, Lux S, Helmstaedter C, Schramm J, Elger
(Moretti Ojemann & Jung, 2006). In addition, we found CE. (2001) Long-term seizure outcome and antiepileptic drug treat-
heterogeneity within surgical and within nonsurgical ment in surgically treated temporal lobe epilepsy patients: a con-
trolled study. Epilepsia 42:1416–1421.
studies in terms of follow-up including outcome mea- Bien CG, Schulze-Bonhage A, Soeder BM, Schramm J, Elger CE,
sures used to determine seizure-freedom, reasons for con- Tiemeier H. (2006) Assessment of the long-term effects of epilepsy
trols not to have surgery, and attrition of patients during surgery with three different reference groups. Epilepsia 47:1865–
1869.
long-term follow-up. Furthermore, there was insufficient Chapell R, Reston J, Snyder D. (2003) Management of treatment-resis-
information to assess the outcome on an intent-treat tant epilepsy. Evidence Report/Technology Assessment No.77.
basis, which would have been desirable given the attri- AHRQ Publication No.03-0028, Rockville, MD. Agency for Health-
care Research and Quality. Part 1. pp. 1–323.
tion rate during long-term observations. Also, we could Cucherat M. (2004) Meta-analysis of surgery trials in refractory epilepsy.
not perform a formal assessment or a ranking of the qual- Rev Neurol (Paris) 160:(5S):232–240. [In French.]
ity of the studies, and some studies had a difference in Derry PA, McLachlan R, Cervinka M. (2001) Surgical versus medical
management of epilepsy: an eight-year prospective outcome analysis
follow-up between the surgical and the nonsurgical arms. of seizure frequency and quality of life. Epilepsia 42(suppl 7):190–
Finally, there may have been a publication bias, as some 191.
of the small studies may not have been published. DerSimonian R, Laird N. (1986) Meta-analysis in clinical trials. Control
Clin Trials 7:177–188.
Although it is important to point out this heterogeneity, Egger M, Schneider M, Smith GD. (1998) Meta-analysis. Spurious preci-
we suggest that a comparison of summary outcome of sion? Meta-analysis of observational studies. BMJ 316: 140–144.
surgery versus nonsurgical treatment may be useful for Engel J, Van Ness PC, Rasmussen TB. (1993) Outcome with respect
to epileptic seizures. In Engel J Jr. (Ed) Surgical treatment of the
guidance and long-term evaluation of the net benefit of epilepsies. Raven Press, New York, pp. 609–621.
surgery that has not been evaluated in a large review. Engel J Jr, Wiebe S, French J, Sperling M, Williamson P, Spencer D,
Finally, we have not included complications of surgery Gumnit R, Zahn C, Westbrook E, Enos B. (2003) Practice parameter:
temporal lobe and localized neocortical resections for epilepsy.
and of AED treatment or effects of withdrawal of AEDs Neurology 60:538–547.
in patients seizure-free after surgery (Schmidt et al., French JA, Kanner AM, Bautista J, Abou-Khalil B, Browne T, Harden
2004) in this review, so we cannot provide a long-term CL, Theodore WH, Bazil C, Stern J, Schachter SC, Bergen D, Hirtz
D, Montouris GD, Nespeca M, Gidal B, Marks WJ Jr, Turk WR,
risk-benefit analysis of surgery versus drug therapy. Fischer JH, Bourgeois B, Wilner A, Faught RE Jr, Sachdeo RC,
The main implications of our results are the following. Beydoun A, Glauser TA. (2004) Efficacy and tolerability of the new
(1) Surgery is four times as likely as medical treatment antiepileptic drugs II: treatment of refractory epilepsy: report of the
Therapeutics and Technology Assessment Subcommittee and Quality
alone to transform drug resistance into full drug response Standards Subcommittee of the American Academy of Neurology
resulting in seizure-free outcome. The difference in effect and the American Epilepsy Society. Neurology 62:1261–1273.

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x
1309
Epilepsy Surgery versus No Surgery

Gilliam F, Kuzniecky R, Meador K, Martin R, Sawrie S, Viikinsalo M, Schiller Y, Najjar Y. (2008) Quantifying the response to antiepileptic
Morawetz R, Faught E. (1999) Patient-oriented outcome assessment after drugs: effect of past treatment history. Neurology 70:54–65.
temporal lobectomy for refractory epilepsy. Neurology 53:662–663. Schmidt D, Baumgartner C, Lçscher W. (2004) Seizure recurrence after
Guldvog B, Loyning Y, Hauglie-Hanssen E, Flood S, Bjornaes H. (1994) planned discontinuation of antiepileptic drugs in seizure-free patients
Surgical treatment for partial epilepsy among Norwegian children after epilepsy surgery: a review of current clinical experience. Epilep-
and adolescents. Epilepsia 35:554–565. sia 45:179–186.
Haglung MM, Moretti Ojemann L. (1993) Seizure outcome in patients Schmidt D, Lçscher W. (2005) Drug resistance in epilepsy: putative
undergoing temporal lobe resections for epilepsy. Neurosurg Clin N neurobiologic and clinical mechanisms. Epilepsia 46:858–877.
Am 4:337–344. Selwa LM, Schmidt SL, Malow BA, Beydoun A. (2003) Long-term out-
Harbord MG, Marson JL. (1987) Temporal lobe epilepsy in childhood: come of nonsurgical candidates with medically refractory localiza-
reappraisal of etiology and outcome. Pediatr Neurol 3:263–268. tion-related epilepsy. Epilepsia 44:1568–1572.
Helmstaedter C, Kurthen MLS, Reuber M, Elger CE. (2003) Chronic epi- Stavem K, Guldvog B. (2005) Long-term survival after epilepsy surgery
lepsy and cognition: a longitudinal study in temporal lobe epilepsy. compared with matched epilepsy controls and the general population.
Ann Neurol 54:425–432. Epilepsy Res 63:67–75.
Higgins JPT, Thompson SG, Decks JJ. (2003) Measuring inconsistency Stavem K, Bjørnaes H, Langmoen IA. (2008) Long-term seizures and
in meta-analyses. Br Med J 327:557–560. quality of life after epilepsy surgery compared with matched controls.
Huttenlocher PR, Hapke RJ. (1990) A follow-up study of intractable Neurosurgery 62: 326–334; discussion 334–335.
seizures in childhood. Ann Neurol 28:699–705. T!llez-Zenteno JF, Dhar R, Wiebe S. (2005) Long-term seizure outcomes
Jones JE, Berven NL, Ramirez L, Woodard A, Hermann BP. (2002) following epilepsy surgery: a systematic review and meta-analysis.
Long-term psychosocial outcomes of anterior temporal lobectomy. Brain 128:1188–1198.
Epilepsia 43:896–903. Tellez-Zenteno JF, Dhar R, Hernandez-Ronquillo L, Wiebe S. (2007)
Kellett MW, Smith DF, Baker GA, Chadwick DW. (1997) Quality of life Long-term outcomes in epilepsy surgery: antiepileptic drugs, mortal-
after epilepsy surgery. J Neurol Neurosurg Psychiatry 63:52–58. ity, cognitive and psychosocial aspects. Brain 130:334–345.
Kumlien E, Doss RC, Gates JR. (2002) Treatment outcome in patients Vickrey BG, Hays RD, Rausch R, Engel J Jr, Visscher BR, Ary CM,
with mesial temporal sclerosis. Seizure 11:413–417. Rogers WH, Brook RH. (1995) Outcomes in 248 patients who had
Kwan P, Brodie MJ. (2000) Early identification of refractory epilepsy. diagnostic evaluations for epilepsy surgery. Lancet 346:1445–1449.
N Engl J Med 342:314–319. Wiebe S, Blume WT, Girvin JP, Eliasziw M. (2001) A randomized, con-
Li H. (2002) Comprehensive therapy in intractable temporal lobe epilep- trolled trial of surgery for temporal-lobe epilepsy. N Engl J Med
sies. A comparison of surgically treated and non-surgically treated 345:311–318.
cases. J Tokyo Med Univ 60:469–482. Wieser HG, Blume WT, Fish D, Goldensohn E, Hufnagel A, King D,
Luciano AL, Shorvon SD. (2007) Results of treatment changes in patients Sperling MR, L#ders H, Pedley TA; Commission on Neurosurgery of
with apparently drug-resistant chronic epilepsy. Ann Neurol 62:375– the ILAE. (2001) Proposal for a new classification of outcome with
381. respect to epileptic seizures following epilepsy surgery. Epilepsia
Markand ON, Salanova V, Whelihan E, Emsley CL. (2000) Health- 42:282–286.
related quality of life outcome in medically refractory epilepsy trea- Yasuda CL, Tedeschi H, Oliveira EL, Ribas GC, Costa AL, Cardoso TA,
ted with anterior temporal lobectomy. Epilepsia 41:749–759. Montenegro MA, Guerreiro CA, Guerreiro MM, Li LM, Cendes F.
McLachlan RS, Rose KJ, Derry PA, Bonnar C, Blume WT, Girvin JP. (2006) Comparison of short-term outcome between surgical and
(1997) Health-related quality of life and seizure control in temporal clinical treatment in temporal lobe epilepsy: a prospective study.
lobe epilepsy. Ann Neurol 41:482–489. Seizure 15:35–40.
Mikati MA, Comair Y, Ismail R, Faour R, Rahi AC. (2004) Effects of
epilepsy surgery on quality of life: a controlled study in a Middle
Eastern population. Epilepsy Behav 5:72–80.
Mikati MA, Comair YG, Rahi A. (2006) Normalization of quality of life Appendix
three years after temporal lobectomy: a controlled study. Epilepsia
47:928–933. Literature search strategy
Moretti Ojemann L, Jung ME. (2006) Outcomes of temporal lobe 1. First step
epilepsy surgery. In Miller JW, Silbergeld DL (Eds) Epilepsy
surgery. Principles and controversies. Taylor & Francis, New York, ‘‘Epilepsy/surgery’’ [MeSH] OR (‘‘epilepsy’’ [TW] AND
pp. 661–698. ‘‘surgery’’ [TW])
O’Donoghue MF, Duncan JS, Sander JW. (1998) The subjective handi-
cap of epilepsy. A new approach to measuring treatment outcome. 2. Second step
Brain 121:317–343.
Ommaya AK. (1963) Behaviour after temporal lobectomy. J Neurol ‘‘Incidence’’ [MESH] OR ‘‘Follow-Up Studies’’ [MESH]
Neurosurg Psychiatry 26:556. OR ‘‘Prognosis’’ [MeSH:NOEXP] OR ‘‘prognos*’’ [TW]
Penfield W, Steelman H. (1947) The treatment of focal epilepsy by corti-
cal excision. Ann Surg 126:740–762.
OR ‘‘predict*’’ [TW] OR ‘‘course’’ [TW] OR ‘‘outcome’’
Penfield W, Paine K. (1955) Results of surgical therapy for focal epilep- [TW] OR ‘‘Survival Analysis’’ [MH:NOEXP]
tic seizures. Can Med Assoc J 73:515–531.
Picot MC, Neveu D, Kahane P, Crespel A, Gelisse P, Hirsch E, Derambu- 3. Third step
re P, Dupont S, Landre E, Chassoux F, Valton L, Vignal JP, Marchal ‘‘randomized controlled trial’’ [PTYP] OR ‘‘random*’’
C, Rougier A, Lamy C, Semah F, Biraben A, Arzimanoglou A,
Petit J, Thomas P, Dujols P, Ryvlin P. (2004) Cost-effectiveness of [TW] OR (‘‘double’’[TW] AND ‘‘blind*’’ [TW]) OR
epilepsy surgery in a cohort of patients with medically intractable partial ‘‘placebo’’ [TW] OR ‘‘drug therapy’’ [SH] OR ‘‘anticon-
epilepsy – preliminary results. Rev Neurol (Paris) 160(5S):354–367. vulsants’’ [SH] OR ‘‘antiepileptic drugs ‘‘ [SH] OR ‘‘thera-
[In French.]
Picot MC, Jaussent A, Kahane P, Crespel A, G!lisse P, Hirsch E, Deram- peutic use’’ [SH:NOEXP] OR ‘‘cohort studies’’ [MESH]
bure P, Dupont S, Landr! E, Chassoux F, Valton L, Vignal JP, Mar- OR ‘‘risk’’ [MESH] OR (‘‘odds’’ [TW] AND ‘‘ratio*’’
chal C, Rougier A, Lamy C, Semah F, Biraben A, Arzimanoglou A, [TW]) OR (‘‘relative’’ [TW] AND ‘‘risk’’ [TW]) OR ‘‘case
Petit J, Thomas P, Neveu D, Ryvlin P. (2008) Medicoeconomic
assessment of epilepsy surgery in adults with medically intractable control*’’ [TW] OR ‘‘case-control studies’’ [MESH]
partial epilepsy. Neurochirurgie 54:484–498. [Article in French.] 4. Limits
Rausch R, Crandall PH. (1982) Psychological status related to surgical
control of temporal lobe seizures. Epilepsia 23:191–202. 1947–2007, Human, Journal article

Epilepsia, 50(6):1301–1309, 2009


doi: 10.1111/j.1528-1167.2008.01997.x

You might also like