West Et Al-2015-Cochrane Database of Systematic Reviews

Cochrane
Library
Cochrane Database of Systematic Reviews

Surgery for epilepsy (Review)
West S, Nolan SJ, Cotton J, Gandhi S, Weston J, Sudan A, Ramirez R, Newton R
West S, Nolan SJ, Cotton J, Gandhi S, Weston J, Sudan A, Ramirez R, Newton R.

Surgery for epilepsy.
Cochrane Database of Systematic Reviews 2015, Issue 7. Art. No.: CD010541.
DOI: 10.1002/14651858.CD010541.pub2.
www.cochranelibrary.com

Copyright © 2015 The Cochrane Collaboration. Published by John Wiley & Sons, Ltd.
Cochrane Trusted evidence.
Informed decisions.

Library Better health. Cochrane Database of Systematic Reviews
TABLE OF CONTENTS
HEADER......................................................................................................................................................................................................... 1
ABSTRACT..................................................................................................................................................................................................... 1
PLAIN LANGUAGE SUMMARY....................................................................................................................................................................... 2
SUMMARY OF FINDINGS.............................................................................................................................................................................. 4
BACKGROUND.............................................................................................................................................................................................. 7
OBJECTIVES.................................................................................................................................................................................................. 8
METHODS..................................................................................................................................................................................................... 8
RESULTS........................................................................................................................................................................................................ 12
Figure 1.................................................................................................................................................................................................. 13
Figure 2.................................................................................................................................................................................................. 21
Figure 3.................................................................................................................................................................................................. 25
DISCUSSION.................................................................................................................................................................................................. 37
AUTHORS' CONCLUSIONS........................................................................................................................................................................... 41
ACKNOWLEDGEMENTS................................................................................................................................................................................ 41
REFERENCES................................................................................................................................................................................................ 42
CHARACTERISTICS OF STUDIES.................................................................................................................................................................. 56
DATA AND ANALYSES.................................................................................................................................................................................... 108
Analysis 1.1. Comparison 1 Surgery versus medical treatment (randomised evidence), Outcome 1 Proportion free from seizure 109
impairing awareness at one year.........................................................................................................................................................
Analysis 1.2. Comparison 1 Surgery versus medical treatment (randomised evidence), Outcome 2 Proportion free from all 109
seizures (including auras) at one year.................................................................................................................................................
Analysis 2.1. Comparison 2 Comparison of surgical techniques (randomised evidence), Outcome 1 Proportion free from seizures 110
(Engel Class 1) at 2 years......................................................................................................................................................................
Analysis 2.2. Comparison 2 Comparison of surgical techniques (randomised evidence), Outcome 2 Proportion free from seizures 110
(Engel Class 1) at 1 year.......................................................................................................................................................................
Analysis 2.3. Comparison 2 Comparison of surgical techniques (randomised evidence), Outcome 3 Proportion free from all 110
seizures (including auras) at 1 year.....................................................................................................................................................
Analysis 3.1. Comparison 3 Surgery for epilepsy (randomised and non-randomised evidence), Outcome 1 Proportion with a 111
good outcome of surgery.....................................................................................................................................................................
Analysis 4.1. Comparison 4 Prognostic factors of a good outcome of surgery (randomised and non-randomised evidence), 118
Outcome 1 Good outcome by MRI results...........................................................................................................................................
Outcome 2 Good outcome by use of intracranial monitoring (IM)....................................................................................................
Outcome 3 Good outcome by presence of mesial temporal sclerosis (MTS)....................................................................................
Outcome 4 Good outcome by Concordance of pre-op MRI and EEG................................................................................................
Outcome 5 Good outcome by history of febrile seizures (FS)............................................................................................................
Outcome 6 Good outcome by history of head injury (HI)..................................................................................................................
Outcome 7 Good outcome by presence of encephalomalacia..........................................................................................................
Outcome 8 Good outcome by presence of focal cortical dysplasia (FCD)/malformation of cortical development (MCD)..............
Outcome 9 Good outcome by presence of tumour............................................................................................................................
Outcome 10 Good outcome by presence of vascular malformation.................................................................................................
Outcome 11 Good outcome by unilateral or bilateral interictal spikes.............................................................................................
Outcome 12 Good Outcome by Extent of Resection..........................................................................................................................
Surgery for epilepsy (Review) i

Informed decisions.

Outcome 13 Good outcome by extent of resection............................................................................................................................
Outcome 14 Good outcome by side of surgical resection.................................................................................................................
Outcome 15 Good outcome by side of surgical resection.................................................................................................................
Outcome 16 Good outcome by presence of postoperative discharges.............................................................................................
Outcome 17 Good outcome by the presence of postoperative discharges.......................................................................................
ADDITIONAL TABLES.................................................................................................................................................................................... 137
APPENDICES................................................................................................................................................................................................. 192
CONTRIBUTIONS OF AUTHORS................................................................................................................................................................... 196
DECLARATIONS OF INTEREST..................................................................................................................................................................... 196
SOURCES OF SUPPORT............................................................................................................................................................................... 196
DIFFERENCES BETWEEN PROTOCOL AND REVIEW.................................................................................................................................... 196
INDEX TERMS............................................................................................................................................................................................... 197
Surgery for epilepsy (Review) ii

Informed decisions.

[Intervention Review]
Surgery for epilepsy
Siobhan West1, Sarah J Nolan2, Jennifer Cotton3, Sacha Gandhi4, Jennifer Weston5, Ajay Sudan1, Roberto Ramirez6, Richard Newton1
1Department of Paediatric Neurology, Royal Manchester Children's Hospital, Manchester, UK. 2Department of Biostatistics, The
University of Liverpool, Liverpool, UK. 3Rheumatology and Orthopaedics, Royal Lancaster Infirmary, University Hospitals of Morecambe
Bay NHS Foundation Trust, Lancaster, UK. 4Undergraduate Medical Education, Manchester Royal Infirmary, Manchester, UK.
5Department of Molecular and Clinical Pharmacology, Institute of Translational Medicine, University of Liverpool, Liverpool, UK. 6Royal
Manchester Children's Hospital, Manchester, UK
Contact address: Siobhan West, Department of Paediatric Neurology, Royal Manchester Children's Hospital, Hathersage Road,
Manchester, M13 0JH, UK. siobhan.west@cmft.nhs.uk, Siobhan.wykes@doctors.net.uk.
Editorial group: Cochrane Epilepsy Group.

Publication status and date: New, published in Issue 7, 2015.
Citation: West S, Nolan SJ, Cotton J, Gandhi S, Weston J, Sudan A, Ramirez R, Newton R. Surgery for epilepsy. Cochrane Database of
Systematic Reviews 2015, Issue 7. Art. No.: CD010541. DOI: 10.1002/14651858.CD010541.pub2.
ABSTRACT
Background
Focal epilepsies are caused by a malfunction of nerve cells localised in one part of one cerebral hemisphere. In studies, estimates of the
number of individuals with focal epilepsy who do not become seizure-free despite optimal drug therapy vary according to the age of
the participants and which focal epilepsies are included, but have been reported as at least 20% and in some studies up to 70%. If the
epileptogenic zone can be located surgical resection offers the chance of a cure with a corresponding increase in quality of life.
Objectives
The primary objective is to assess the overall outcome of epilepsy surgery according to evidence from randomised controlled trials.
The secondary objectives are to assess the overall outcome of epilepsy surgery according to non-randomised evidence and to identify the
factors that correlate to remission of seizures postoperatively.
Search methods
We searched the Cochrane Epilepsy Group Specialised Register (June 2013), the Cochrane Central Register of Controlled Trials (CENTRAL
2013, Issue 6), MEDLINE (Ovid) (2001 to 4 July 2013), ClinicalTrials.gov and the World Health Organization (WHO) International Clinical Trials
Registry Platform (ICTRP) for relevant trials up to 4 July 2013.
Selection criteria
Eligible studies were randomised controlled trials (RCTs), cohort studies or case series, with either a prospective and/or retrospective
design, including at least 30 participants, a well-defined population (age, sex, seizure type/frequency, duration of epilepsy, aetiology,
magnetic resonance imaging (MRI) diagnosis, surgical findings), an MRI performed in at least 90% of cases and an expected duration of
follow-up of at least one year, and reporting an outcome relating to postoperative seizure control.
Data collection and analysis

Three groups of two review authors independently screened all references for eligibility, assessed study quality and risk of bias, and
extracted data. Outcomes were proportion of participants achieving a good outcome according to the presence or absence of each
prognostic factor of interest. We intended to combine data with risk ratios (RR) and 95% confidence intervals.
Surgery for epilepsy (Review) 1

Informed decisions.

Main results
We identified 177 studies (16,253 participants) investigating the outcome of surgery for epilepsy. Four studies were RCTs (including one
that randomised participants to surgery or medical treatment). The risk of bias in the RCTs was unclear or high, limiting our confidence in
the evidence that addressed the primary review objective. Most of the remaining 173 non-randomised studies had a retrospective design;
they were of variable size, were conducted in a range of countries, recruited a wide demographic range of participants, used a wide range
of surgical techniques and used different scales used to measure outcomes. We performed quality assessment using the Effective Public
Health Practice Project (EPHPP) tool and determined that most studies provided moderate or weak evidence. For 29 studies reporting
multivariate analyses we used the Quality in Prognostic Studies (QUIPS) tool and determined that very few studies were at low risk of bias
across the domains.
In terms of freedom from seizures, one RCT found surgery to be superior to medical treatment, two RCTs found no statistically significant
difference between anterior temporal lobectomy (ATL) with or without corpus callosotomy or between 2.5 cm or 3.5 cm ATL resection,
and one RCT found total hippocampectomy to be superior to partial hippocampectomy. We judged the evidence from the four RCTs to
be of moderate to very low quality due to the lack of information reported about the randomised trial design and the restricted study
populations.
Of the 16,253 participants included in this review, 10,518 (65%) achieved a good outcome from surgery; this ranged across studies from
13.5% to 92.5%. Overall, we found the quality of data in relation to the recording of adverse events to be very poor.
In total, 118 studies examined between one and eight prognostic factors in univariate analysis. We found the following prognostic factors
to be associated with a better post-surgical seizure outcome: an abnormal pre-operative MRI, no use of intracranial monitoring, complete
surgical resection, presence of mesial temporal sclerosis, concordance of pre-operative MRI and electroencephalography (EEG), history
of febrile seizures, absence of focal cortical dysplasia/malformation of cortical development, presence of tumour, right-sided resection
and presence of unilateral interictal spikes. We found no evidence that history of head injury, presence of encephalomalacia, presence
of vascular malformation or presence of postoperative discharges were prognostic factors of outcome. We observed variability between
studies for many of our analyses, likely due to the small study sizes with unbalanced group sizes, variation in the definition of seizure
outcome, definition of the prognostic factor and the influence of the site of surgery, all of which we observed to be related to postoperative
seizure outcome. Twenty-nine studies reported multivariable models of prognostic factors and the direction of association of factors with
outcome was generally the same as found in the univariate analyses. However, due to the different multivariable analysis approaches and
selective reporting of results, meaningful comparison of multivariate analysis with univariate meta-analysis is difficult.
Authors' conclusions
The study design issues and limited information presented in the included studies mean that our results provide limited evidence to aid
patient selection for surgery and prediction of likely surgical outcome. Future research should be of high quality, have a prospective design,
be appropriately powered and focus on specific issues related to diagnostic tools, the site-specific surgical approach and other issues
such as the extent of resection. Prognostic factors related to the outcome of surgery should be investigated via multivariable statistical
regression modelling, where variables are selected for modelling according to clinical relevance and all numerical results of the prognostic
models are fully reported. Protocols should include pre- and postoperative measures of speech and language function, cognition and
social functioning along with a mental state assessment. Journal editors should not accept papers where adverse events from a medical
intervention are not recorded. Improvements in the development of cancer care over the past three to four decades have been achieved by
answering well-defined questions through the conduct of focused RCTs in a step-wise fashion. The same approach to surgery for epilepsy
is required.
PLAIN LANGUAGE SUMMARY
Surgery for epilepsy
Focal epilepsies are caused by abnormal electrical discharges in specific (localised) parts of the brain. In most people the resulting epileptic
seizures can be controlled with medication. In up to 30% of people these seizures are not controlled by medication. If the site of origin
of these signals (the epileptogenic zone) can be located from the description of the seizures, or from magnetic resonance imaging (MRI)
(a medical imaging scan that uses strong magnetic fields and radio waves to produce detailed images of the inside of the body) and
electroencephalography (EEG) findings (recording of electrical activity along the scalp) the person should be offered the chance of having
the epileptogenic zone removed. We studied the factors (characteristics of the people undergoing surgery and details of surgery type) that
might be linked to the best chance of surgical cure of epileptic seizures.
We examined evidence from 177 included studies reporting the experience of 16,253 people of all ages. In total, 10,518 people (65% of the
total who had surgery in all studies) experienced a good outcome from surgery, defined as freedom from epileptic seizures.
One randomised controlled trial (RCT) established the superiority of surgery over continuing to try different medication in drug-resistant
(intractable) epilepsy for the individuals in this trial. Three RCTs compared different types of surgery. No difference in seizure outcome was
found in two trials; one removing 2.5 cm or 3.5 cm of the anterior temporal lobe (ATL, a part of the brain in which the epileptogenic zone
is often located) or surgically removing the ATL with or without an additional procedure to sever the nerves that connect the two halves

Informed decisions.

(hemispheres) of the brain. In the third trial, completely removing the hippocampus (a part of the brain in which the epileptogenic zone is
often located) was found to be superior to only removing part of the hippocampus.
Most of the studies in this review were poor quality and had a retrospective design (a design where individuals are recruited after the
result of surgery has been recorded, which looks back for the existence of factors related to the result of surgery). They used variable
surgical approaches for different sites of the brain, different processes to select candidates for surgery, different definitions of freedom
from seizures after surgery and measured these outcomes at varying points. Fewer than half the studies gave details of complications and
deaths associated with surgery.
We identified some factors that predicted a better outcome from surgery. These were: a well-defined lesion (abnormality) on the MRI scan
that corresponded with what was expected from the description of the seizures and the EEG findings, complete surgical removal of the
lesion and a history of febrile seizures (a seizure associated with fever in a young child) often associated with mesial temporal sclerosis
(scarring in the inner portions of the temporal lobe of the brain). More diffuse (spread out) brain abnormalities that might be associated with
brain injury (from trauma or otherwise) or an abnormality of brain development were not associated with a good outcome. The presence of
such abnormalities is often associated with a need to embark on more detailed pre-operative investigations including intracranial (inside
the skull) EEG monitoring. We would have liked to examine the collective effect of these factors (i.e. what would be the effect on outcome
if a person has a history of febrile seizures, trauma and an MRI lesion altogether), however the studies did not report enough information
to allow this.
We recommend that future studies should have a prospective design (a design where individuals are recruited before surgery has taken
place, which identifies factors of interest before surgery and follows up individuals after surgery to record the outcome). Studies should
examine the collective effect of factors that may predict the outcome of surgery using appropriate statistical methods. Future studies
should use pre-operative investigations to guide the selection of the best candidates for surgery and the best surgical approach, use clearly
defined methods to measure the result of surgery at specific time points and include pre- and postoperative cognitive and mental state
assessment. Deaths during or after surgery, and complications and side effects from surgery, should be clearly recorded.

SUMMARY OF FINDINGS

Summary of findings for the main comparison.
Library
Cochrane
Surgery compared with medical treatment for epilepsy
Patient or population: adults with drug-resistant epilepsy suitable for surgical intervention
Settings: outpatients (following surgery in hospital)
Better health.
Informed decisions.
Trusted evidence.
Intervention: surgery
Comparison: medical treatment
Outcomes Illustrative comparative risks* (95% CI) Relative effect No of par- Quality of the Comments
(95% CI) ticipants evidence
Assumed risk1 Corresponding risk2 (studies) (GRADE)
Medical treat- Surgery

ment
Proportion free from seizure im- 75 per 1000 500 more per 1000 RR 7.67 (2.50 to 80 (1 study) ⊕⊝⊝⊝ RR > 1 indicates ad-
pairing awareness at 1 year (113 to 925 more) 23.51) very low,3,4,5 vantage
for surgery
Proportion free from all seizures 23 per 1000 350 per 1000 RR 15.00 (2.08 to 80 (1 study) ⊕⊝⊝⊝ RR > 1 indicates ad-
(including auras) at 1 year (27 to 973 more) 108.23) very low,3,4,5 vantage
for surgery
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is
based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: confidence interval; RR: risk ratio

GRADE Working Group grades of evidence
High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Very low quality: We are very uncertain about the estimate.
1The assumed risk is the event rate in the control group (medical treatment).
2Upper bounds of corresponding risk interval revised to their maximum to align with the upper bound of the confidence interval of the relative effect.
3Large but imprecise effect size shown in favour of surgical treatment (downgraded due to imprecision as relatively small study and low event rate in the control group).
4

4Downgraded due to insufficient information regarding methods of randomisation and allocation concealment in the study (see Table 4).
5Downgraded for indirectness: results are applicable to adults (over 16 years only), children excluded from the study.

Library
Cochrane

Summary of findings 2.
Comparison of surgical interventions for epilepsy
Patient or population: adults and children with drug-resistant epilepsy suitable for surgical intervention
Better health.
Informed decisions.
Trusted evidence.
Settings: outpatients (following surgery in hospital)
Intervention: experimental surgical intervention (see comments)
Comparison: control surgical intervention (see comments)
Outcomes Illustrative comparative risks* (95% CI) Relative No of Qual- Comments

effect partici- ity of
Assumed Corresponding risk (95% pants the evi-
risk1 CI) (stud- dence
ies) (GRADE)
Control Experimental
surgical in- surgical intervention

tervention
(see comments)
(see com-
ments)
Proportion free from 600 per 132 more per 1000 RR 1.22 60 ⊕⊕⊕⊝ Experimental Intervention is anterior temporal lobectomy with
seizures (Engel Class 1000 (90 less to 400 more)2 moder- corpus callosotomy (aCCT)
1) at 2 years (0.85 to (1 study) ate3
1.76) Control intervention is anterior temporal lobectomy
without corpus callosotomy (ATL)

RR > 1 indicates advantage for aCCT
Proportion free from 728 per 15 more per 1000 RR 1.02 207 ⊕⊕⊕⊝ Experimental Intervention is a 2.5 cm resection
seizures (Engel Class 1000 (102 less to 146 more) moder-
1) at 1 year (0.86 to (1 study) ate4,5 Control intervention is 3.5 cm resection
1.2)
RR > 1 indicates advantage for 2.5 cm resection
Proportion free from 382 per 314 per 1000 RR 1.82 70 ⊕⊕⊝⊝ Experimental intervention is total resection
all seizures (including 1000 (46 more to 718 more)2 low5,6
auras) at 1 year (1.12 to (1 study) Control intervention is partial resection
2.93)
5

RR > 1 indicates advantage for total resection
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is
Library
Cochrane
based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI).
CI: confidence interval; RR: risk ratio
GRADE Working Group grades of evidence

High quality: Further research is very unlikely to change our confidence in the estimate of effect.
Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate.
Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate.
Better health.
Informed decisions.
Trusted evidence.
Very low quality: We are very uncertain about the estimate.
1The assumed risk is the event rate in the control group (medical treatment).
2Upper bounds of corresponding risk interval revised to their maximum to align with the upper bound of the confidence interval of the relative effect.
3Inadequate method of quasi-randomisation used (allocation based on odd and even patient ID numbers) and unclear if participants/personnel/outcome assessors were blinded
(see Table 4).
4Study judged to be a low risk of bias for all domains (see Table 4).
5Downgraded for indirectness: results are applicable to adults (over 18 years only), children excluded from the study.
6Downgraded due to insufficient information regarding methods of randomisation and allocation concealment in the study (see Table 4).


6

Informed decisions.

BACKGROUND complex in short-term memory, consciousness and emotions. As

resection of the anteromesial structures became the accepted
Description of the condition treatment for temporal lobe epilepsy the hippocampus was
elucidated as important for short-term recent memory function.
Epilepsy has very recently been redefined by the International
This observation led to the incorporation of a neuropsychologist
League Against Epilepsy as "a disease of the brain defined by
in most surgery selection teams. The temporary and partial
any of the following conditions: (1) At least two unprovoked (or
suppression of one cerebral hemisphere by injection of intracarotid
reflex) seizures occurring > 24 h apart; (2) one unprovoked (or
sodium amytal, a technique introduced by Wada, became a useful
reflex) seizure and a probability of further seizures similar to
test for determining the laterality of speech function and the
the general recurrence risk (at least 60%) after two unprovoked
evaluation of memory responses in people with bitemporal seizure
seizures, occurring over the next 10 years; (3) diagnosis of an
activity. Functional magnetic resonance imaging (MRI), along with
epilepsy syndrome" (Fisher 2014). It is a common condition
neuropsychometry, is now superceding this initial approach.
with a prevalence of around 1 in 200 people. Despite optimal
pharmacotherapy, about 20% to 30% of individuals do not become Other sophisticated technological developments have followed:
seizure-free (Annegers 1979; Collaborative 1992; Cockerell 1995; the advent of computerised tomography in the 1970s; magnetic
Kwan 2000). For some of these people, surgery is a therapeutic resonance imaging in the 1990s; computerised analysis of ictal
option. and inter-ictal EEG activity; functional MRI with psychometric
analysis and ever more sophisticated stereotaxis guiding the
Description of the intervention placement of deep electrodes for long-term EEG analysis and
The intervention involves the localisation of the epileptogenic surgical intervention. Most recently these techniques have been
focus and then, if the potential benefit is assessed to outweigh complemented by the co-registration of single photon emission
the risk, it is resected surgically. The history of the development computed tomography (SPECT) and positron emission tomography
of surgery for epilepsy has been well reviewed by Feindel 2009. (PET) findings (Feindel 2009). These developments are now leading
This work was pioneered by Victor Horsley when, in 1886, he to a more precise localisation of epileptogenic foci and a reduction
operated on a 22-year old man who had developed a focal epilepsy in the risk of removing eloquent cortex. This approach has led
following a head injury. A highly vascular scar associated with an to more opportunity for the accurate assessment of any person
old depressed comminuted skull fracture was excised along with a with drug-resistant focal epilepsy for a surgical cure. Sophisticated
border of cortex. The presence of a discrete cortical vascular scar technology only has a place, however, in the setting of a good
and the arrest of the focal motor seizures following its excision inter-disciplinary team working in harmony and incorporating
provided direct support for Hughlings Jackson's concept of the the skills of a neurologist, neurophysiologist, psychiatrist,
aetiology of focal epilepsy. It subsequently became evident that neuropsychologist and neurosurgeon with postoperative help
a variety of pathologies could give rise to the focal epilepsies. from remedial therapists who have good scope for liaison with
Techniques of cortical stimulation along with the advent of educational, vocational and social services for good postoperative
electroencephalography in the 1930s aided localisation. Until the rehabilitation.
1940s surgery was directed mainly to the convexity of the cerebral
hemispheres and most often for removal of traumatic scars or Success rates for resective epilepsy surgery are estimated to
tumours. The work of Frederic Gibbs and William Lennox from have increased from 43% to 85% during the period 1986 to
1936 promoted EEG into a strategic position for diagnosis and 1999 (Engel 1993a; Engel 2003; National 1990a). Data from
early classification of the epilepsies. Herbert Jasper, working with multiple sources suggest that 55% to 70% of individuals
Wilder Penfield, used EEG to develop new approaches for surgery undergoing temporal resection and 30% to 50% of individuals
for epilepsy particularly in the temporal lobe and mesial temporal undergoing extratemporal resection become completely seizure-
structures. A major obstacle to the removal of these structures was free. A prospective randomised controlled trial of surgery for
the lack of knowledge about their function (Feindel 2009). temporal lobe epilepsy showed that 58% of individuals randomised
to surgery were seizure-free compared to 8% of the medical group
Penfield considered EEG and emerging techniques for (Wiebe 2001). Surgery is considered a valuable option for medically
electrocorticography (ECoG) useful if they could disclose intractable epilepsy, even in the absence of proven drug resistance
pathological areas in the brain. If a visible lesion was not (Engel 1993b).
found at the place indicated by the recording of epileptic
activity, Penfield would usually decline to perform a resection. How the intervention might work
A success rate of just over 50% indicated that resection limited The rationale for the intervention is the initial localisation of
to the anterolateral temporal cortex did not eliminate all the the epileptogenic focus and then its surgical resection. The
epileptogenic tissue in many people. Greater interest was then mainstay for investigation is magnetic resonance imaging (MRI).
paid to mesial temporal structures. Jasper noted that a cure could Concordance between an MRI scan and EEG findings along with
be effected even when there was no visible abnormality in the seizure semiology is sought. The key is the accurate localisation of
excised material and added, "it seems clear, therefore, that the the epileptogenic focus to an area of the brain that might safely be
pathophysiological state of spike foci may not always be associated removed without inducing a neurological impairment. If the lesion
with structural alterations which can be seen by present methods is not well defined then further imaging using PET or SPECT may
of microscopic examination." (Feindel 2009). This led to more supplement the accurate placement of indwelling EEG electrodes
detailed pathological study of excised material, an approach led by to achieve this aim (see Description of the intervention). Where the
Murray Falconer with Alfred Meyer. More sophisticated EEG study, epileptogenic focus can safely be removed the epilepsy may be
cortical stimulation and detailed descriptions of seizure semiology cured, with a corresponding improvement in the quality of life.
identified the importance of the human claustroamygdaloid

Informed decisions.

Why it is important to do this review 3. A well-defined population (age, sex, seizure type and frequency,
duration of epilepsy, aetiology, magnetic resonance imaging
Surgical outcomes may be greatly influenced by the presence of (MRI) diagnosis, surgical findings).
selected prognostic indicators (Berg 1998; Tonini 1997). However,
4. MRI performed in at least 90% of cases.
there are still uncertainties about which patients are most likely
to achieve good surgical outcomes. Good surgical outcomes 5. Expected duration of follow-up of at least one year.
appear to be associated with a number of factors (hippocampal 6. An outcome relating to postoperative seizure control is reported.
sclerosis, anterior temporal localisation of interictal epileptiform
activity, absence of pre-operative generalised seizures and absence We excluded reports if they were in abstract form, in book chapters
of seizures in the first postoperative week) (McIntosh 2001). or if they were not sufficiently clear about their methods, if they
However, the published trial results are frequently confusing and were written in languages other than English, Italian, French,
contradictory, thus preventing inferences for clinical practice. German or Spanish (due to the availability of translators for detailed
This is the first Cochrane review to investigate the association data extraction), or if they did not meet all the above inclusion
between specific prognostic factors and surgical outcome. It will criteria. We also excluded repeated publications from the same
complement and update the only systematic review to date to institution (among which only the most recent was retained for
examine factors predictive of the outcome of epilepsy surgery review) unless they dealt with different prognostic factors.
(Tonini 2004). It will inform the surgical selection process and allow
Types of participants
the refinement of the risk/benefit analysis for surgical intervention.
Children, adolescents and adults considered surgical candidates,
OBJECTIVES having drug-resistant partial seizures and secondarily generalised
seizures of temporal or extratemporal origin, i.e. seizures that
The primary objective is to assess the overall outcome of epilepsy continue despite treatment with anticonvulsant medication.
surgery according to evidence from randomised controlled trials.
Types of interventions
The secondary objectives are to assess the overall outcome
of epilepsy surgery according to non-randomised evidence and Surgical treatment for drug-resistant partial seizures and
to identify the factors that correlate to remission of seizures secondarily generalised seizures of temporal or extratemporal
postoperatively. origin. For RCTs we considered all control groups, such as
medical treatment or no treatment, or comparisons of different
METHODS surgical techniques. We considered non-randomised studies with
or without control groups (i.e. case series) for inclusion.
Criteria for considering studies for this review
Types of outcome measures
Types of studies
Primary outcomes
We included studies if they satisfied the following criteria:
Seizure outcome (proportion achieving a good outcome from surgery)
1. Randomised controlled trial (RCT), cohort study or case series,
The outcome of seizures after epilepsy surgery is classified
either prospective and/or retrospective.
according to Engel's four categories with subcategories (Engel 1987;
2. A sample size of at least 30 patients. Engel 1993b), or it is reported as such when different definitions are
used (see Table).

Engel class Description (subclasses)
Class 1: Free of disabling 1A: Completely seizure-free since surgery

seizures 1B: Non-disabling simple partial seizures only since surgery
1C: Some disabling seizures after surgery, but free of disabling seizures for at least 2 years
1D: Generalised convulsion with antiepileptic drug withdrawal only Class 2 (rare disabling seizures
'almost seizure-free')
Class 2: Almost seizure-free 2A: Initially free of disabling seizures but has rare seizures now
(rare disabling seizures) 2B: Rare disabling seizures since surgery
2C: More than rare disabling seizures after surgery, but rare seizures for at least 2 years
2D: Nocturnal seizures only
Class 3: Worthwhile improve- 3A: Worthwhile seizure reduction

ment 3B: Prolonged seizure-free intervals amounting to greater than half the follow-up period, but not
less than 2 years
Class 4: No worthwhile im- 4A: No significant seizure reduction

provement 4B: No appreciable change
4C: Seizures worse

Informed decisions.


Sources: Engel 1987; Engel 1993b. • History of head injury: yes or no.
• Encephalomalacia on pathology: present or absent.
We considered:
• Focal cortical dysplasia/malformation of cortical development
• 'good outcome' as seizure control or seizure-free status for at on pathology: present or absent.
least one year or Engel class 1 (where individual study data • Tumour on pathology: present or absent.
quality did not allow further refinement);
We included studies that referred to pathological results of
• 'improved outcome' as near complete control or moderate
'tumour' or where specific tumour types were reported.
improvement or Engel classes 2 and 3;
• 'worse outcome' as slightly reduced or unchanged or worsened • Vascular malformation on pathology: present or absent.
seizure frequency or Engel class 4.
We included studies that referred to pathological results of
We also considered, where data for these time points were 'vascular malformation' or where specific types of vascular
available, results at 12 and 24 months. Reporting of this primary malformations were reported.
outcome using the outcome scales described above was not an
eligibility requirement for inclusion in this review. However, we • Distribution of interictal spikes: unilateral or bilateral.
excluded studies that did not report an outcome relating to seizure
control following surgery. We considered other outcome scales that Our interest was in whether interictal spikes (epileptiform EEG
satisfied our above definitions. We divided studies into subgroups discharges noted between seizures) were related to the area to be
based on seizure outcome defined by Engel Class Scale, more than excised at surgery or whether they were more widespread. Terms
one year seizure-free or other scale (see Subgroup analysis and also used in the studies are: lateralising versus non-lateralising
investigation of heterogeneity). spikes (that is, discharges on the side to be operated or on both
sides of the brain); focal versus non-focal spikes or localising versus
For the purposes of this review, we compared a 'good non-localising spikes (that is, discharges seen only related to the
outcome' (seizure remission as defined above) with a 'poor surgical site or seen to be more widespread).
outcome' where a poor outcome is defined as the improved and
worse outcome categories as above combined (i.e. Engel class 2 to Operative factors
4 or not seizure-free for at least one year). We did not consider other • Extent of surgical resection: complete or incomplete.
combinations of the outcome scales. For trials that reported other
combinations of the outcome scales, or used a scale that did not We anticipated that the definition of a 'complete' or 'less complete'
clearly satisfy our definitions, where possible we contacted the trial resection would be variable across studies. The definition for
authors to request further information about seizure outcome data. most was defined by the type of surgery performed; e.g. anterior
Where further information could not be provided, we excluded temporal lobectomy or extended resection is complete resection
studies from analysis in the review (see Data synthesis), but we while selective amygdalohippocampectomy or lesionectomy is less
retained them in the narrative section of the review. complete resection. Where other clear descriptions were given,
we included these studies: e.g. postoperative MRI appearance;
Secondary outcomes intraoperative subdural EEG findings; intra-operative surgical
Seizure outcome according to prognostic factors of interest description or dimensions of resected areas.
We considered the proportion of individuals with a good outcome • Side of surgical resection: left-sided or right-sided resection
from surgery (see Primary outcomes) according to the following
prognostic factors that we considered to be of clinical relevance: Postoperative factors
• Postoperative discharges: presence or absence of EEG

Pre-operative factors
epileptiform discharges in the postoperative period.
• Results of pre-operative MRI: normal (i.e. no visible abnormality
present on MRI) or abnormal (i.e. visible abnormality present on We dichotomised all factors according to the definitions above for
MRI). analysis. We included data reported according to our definitions
above or where data were reported in a way that allowed us to
We included studies where MRI results were referred to only as categorise into the above definitions (e.g. if specific MRI results
'abnormal' and where specific abnormalities were reported. were reported for all individuals, we categorised them into 'normal'
and 'abnormal' or 'concordant' and 'discordant' with EEG results
• Use of pre-operative intracranial (invasive) monitoring: yes or etc.) We considered other definitions reported in the included
no. studies if equivalent (or approximately equivalent) to our pre-
• Mesial temporal sclerosis (MTS) on MRI or pathology: present or specified definitions.
absent.
• Concordance of pre-operative MRI and EEG: yes or no. Search methods for identification of studies
A MEDLINE (OVID) search was carried out for Tonini 2004 to identify
Concordance relates to whether EEG discharges arise from the
relevant studies published between 1984 and 2001, where 1984
same area of the brain identified as abnormal on the MRI scan (i.e.
coincides with the introduction of magnetic resonance imaging
the surgically targeted area).
(MRI). We used the results found in that review and carried out
• History of febrile seizures: yes or no. searches to cover 2001 onwards.

Informed decisions.

Electronic searches 6. prognostic indicators: different indicators are described as

factors affecting the outcome of epilepsy surgery in terms
We searched:
of seizure remission (specifically focal cortical dysplasia/
1. Cochrane Epilepsy Group Specialised Register (June 2013) using malformation of cortical developments, febrile seizures,
the search strategy outlined in Appendix 1; tumours, vascular disorders, CNS infections, mesial temporal
sclerosis, abnormal MRI, electroencephalogram (EEG)/MRI
2. Cochrane Central Register of Controlled Trials (CENTRAL 2013,
concordance, interictal spikes, intracranial monitoring, extent
Issue 6) using the search strategy in outlined in Appendix 2;
of resection, postoperative discharges, other factors studied).
3. MEDLINE (Ovid) (2001 to 4 July 2013) using the search strategy Mesial temporal sclerosis and tumours require pathological
outlined in Appendix 3; confirmation. We also recorded details of statistical analysis of
4. ClinicalTrial.gov (4 July 2013) using the search terms 'surgery prognostic indicators and multivariable prognostic models (if
and epilepsy'; reported).
5. World Health Organization (WHO) International Clinical Trials
Registry Platform (ICTRP) (4 July 2013) using the search terms The review was limited to prognostic factors that are clinically
'epilepsy' in the condition AND 'surgery' in the intervention. relevant (detailed in Secondary outcomes) and/or reported by at
least two studies.
We imposed language restrictions due to the availability of
translators for extensive data extraction (English, Italian, French, Consensus is also required on each variable reported in the data
German or Spanish). There is no Cochrane approved or collection form; any disagreement led to a discussion of the issue by
recommended search filter for non-randomised studies, so in the the two review authors with any persisting disagreement resolved
MEDLINE search strategy the Cochrane Epilepsy Group's Trials by an independent third author. In selected cases, an independent
Search Co-ordinators chose the terms for the required types of non- evaluator (SN) resolved conflicting data.
randomised studies. We did not subject the resulting search filter to
Assessment of risk of bias in included studies
any systematic testing before use.
For RCT evidence, we assessed all domains of the current Cochrane
Searching other resources 'Risk of bias' tool (Higgins 2011).
We also examined the reference lists of included studies for further
For non-randomised evidence, we employed the Effective Public
relevant studies for inclusion in this review.
Health Practice Project (EPHPP) tool, which is appropriate for case
Data collection and analysis series study designs. See Appendix 4.
Selection of studies Two review authors (SW and RN) assessed risk of bias and
two review authors (SN and JW) independently checked the
Two review authors (SW and RN) independently assessed trials judgements.
for inclusion. We resolved any disagreements through mutual
discussion and sought the opinion of a third author (SN) where We planned to incorporate 'Risk of bias' assessments into the
necessary. analysis if deemed appropriate, using sensitivity analysis, in that a
secondary analysis of the data would include only studies rated as
Data extraction and management low in quality with results presented in the Results section of the
We implemented a database search for the identification and review. However, following use of the EPHPP tool, we found this
inclusion of relevant articles, i.e. those that fulfil the inclusion tool to be inadequate to judge the relative quality of the included
criteria and have complete information about the outcome of studies so we concluded that incorporation of quality assessment
epilepsy and prognostic factors. into the analysis would be inappropriate (see Sensitivity analysis
and Risk of bias in included studies for further details).
SW, RN, JC, AS and SG collected the data using a semi-structured
form via a Microsoft Access database (created by SN) for each study. Post hoc review of multivariable prognostic models: SN assessed
RR joined RN to extract data from the Spanish papers. risk of bias according the to Quality in Prognosis Studies (QUIPS)
tool (Hayden 2006).
We considered the following variables:
We presented the primary outcome (seizure outcome) for studies
1. methods of assessment of eligible studies; with a randomised controlled design in 'Summary of findings'
2. demographic and clinical characteristics (number of patients tables and we judged the quality of the evidence contributing to
selected for surgery, age (with special attention to patients these outcomes according to GRADE (Grading of Recommendations
younger and older than 12 years), sex, disease duration, history Assessment, Development and Evaluation) criteria (Guyatt 2008).
of febrile seizures or relevant central nervous system (CNS)
disorder); Measures of treatment effect
3. MRI pre-operative diagnosis (mesial temporal sclerosis, We measured the outcome of seizures after epilepsy surgery
tumours, other CNS abnormalities, normal); as good compared to poor outcome overall and according to
4. surgical findings (age at surgery, side of resection, surgical presence or absence of prognostic factors of interest in univariate
procedure (temporal or extratemporal), extent of resection); analysis (see Types of outcome measures and Data synthesis).
5. histopathological diagnosis (same categories as MRI); duration We analysed all outcomes as dichotomous outcomes summarised
of follow-up; post-surgery findings (drop-outs, adverse events); with risk ratios (RR) with 95% confidence intervals. We considered

Informed decisions.

multivariable prognostic models narratively as reported in the of adjusted and unadjusted estimates and compare results.
original study publications. However, insufficient adjusted data were presented to allow us
to perform meta-analysis of adjusted results from multivariable
Unit of analysis issues prognostic models, therefore we summarised all multivariable
We did not encounter any unit of analysis issues. The unit of models narratively and comparisons of multivariable adjustments
intervention and analysis was the individual for all included studies to univariate analyses were narrative.
and no studies were of a repeated measures (longitudinal) nature
Where we found considerable statistical heterogeneity to be
or of a cross-over design.
present (Chi2 test for heterogeneity P value < 0.1 and/or I2 greater
Dealing with missing data than 50%), we used a random-effects meta-analysis and performed
subgroup and sensitivity analyses investigating differences in study
We attempted to seek missing statistics from studies through characteristics and participant factors.
contact with the study authors. In the case of missing data, we
attempted to clarify the reasons for missing data to determine if Subgroup analysis and investigation of heterogeneity
data were missing at random or not. We analysed all data according
In the case of a substantial amount of heterogeneity being found
to the intention-to-treat principle.
across univariate prognostic factors, we performed further analyses
Assessment of heterogeneity such as stratification, subgroup analyses and sensitivity analyses to
examine the differences in study characteristics (such as outcome
We assessed the existence of clinical heterogeneity by and prognostic factor definition, study design and quality) and
examining the differences in study characteristics and participant participant demographic factors.
demographic factors in order to inform decisions regarding the
combination of study data. We assessed statistical heterogeneity Good and poor outcomes were measured using varying definitions
by visually inspecting forest plots and using a Chi2 test for across studies; i.e. a good outcome can be defined as seizure-
heterogeneity (with a P value of 0.10 for significance) and the free status for at least one year, Engel class 1 or other equivalent
I2 statistic as a measure of inconsistency across studies, with definitions. We defined these scales as Engel Class Scale, more than
an I2 value of 50% to 75% or higher representing substantial one year seizure-free and other scale; see Primary outcomes for
heterogeneity (Higgins 2011). Where we found considerable further details. The effect of the variation of outcome definition
statistical heterogeneity according to the I2 statistic value (> 50%), wasto be examined by subgroup analysis.
we performed meta-analysis with a random-effects model rather
Due to differences in surgical technique and/or associated
than a fixed-effect model, in addition to subgroup and sensitivity
pathology with location of surgery (temporal or extratemporal
analyses investigating differences in study characteristics and
lobe), where applicable we also performed stratified analyses,
participant factors.
grouping studies into the following categories: all participants
Assessment of reporting biases in the study had temporal lobe surgery (temporal lobe);
all participants in the study had extratemporal lobe surgery
To enable comparison of outcomes of interest, we would require (extratemporal lobe); study included a combination of participants
all protocols from study authors. However, due to the large number with temporal lobe and extratemporal lobe surgery (combination).
of included studies in this review, obtaining all protocols was
impractical and impossible, so we made a judgement on the Sensitivity analysis
existence of reporting bias. If we suspected reporting bias, we
We intended to perform sensitivity or subgroup analyses to
investigated further using the ORBIT classification system (Kirkham
examine the effect of study quality based on the 'Risk of bias' and
2010).
quality assessment tools. However, due to the inadequacy of the
We examined publication bias by identifying unpublished data quality assessment tool used in separating studies of generally poor
by carrying out a comprehensive search of multiple sources and methodological design (majority retrospective case series), we
requesting any unpublished data from authors. We looked for decided that sensitivity or subgroup analysis according to 'quality
small-study effects to establish the likelihood of publication bias assessment' would not be informative or appropriate (see Risk of
and we examined asymmetry of funnel plots. bias in included studies for more information).
Data synthesis Instead, as described in Subgroup analysis and investigation of

heterogeneity, we considered differences in study characteristics
For each prognostic factor of interest individually (see point and participant demographic factors as sources of heterogeneity
six of Data extraction and management), we performed a in analyses. Further, in the case of an extreme result reported by a
univariate aggregate data fixed-effect meta-analysis using the study (a particularly large effect in favour of the presence of absence
Mantel-Haenszel method to assess the presence or absence of of a prognostic factor), we double-checked the extracted data on
that factor as an independent predictor of the outcome of surgery this factor from the study publication and investigated study or
(good or poor outcome), analysed as a dichotomous outcome and participant-related characteristics that could have contributed to
presented as a pooled risk ratio (RR) and 95% confidence interval. the large effect.
Post-hoc analysis: We also investigated whether the effect of
other prognostic factors on any individual prognostic factor had
also been adjusted for (for example, in multivariable regression
models). In this case, we hoped to perform separate meta-analyses

Informed decisions.

RESULTS From searches conducted by the Cochrane Epilepsy Group between

2001 and 2010 we identified 833 records. From searches conducted
Description of studies after the initiation of this review in 2010, we identified 575 records.
All searches from 2001 were conducted in the databases using the
Results of the search
search strategies outlined in Electronic searches. We removed 96
A MEDLINE search conducted over the years of 1984 to 2001 duplicate records from those identified between 2010 and 2013
identified 1051 records, 619 of which we excluded as 'prognosis and screened 1395 records (title and abstract) for inclusion in
of seizures after surgery was not measured' (in other words no the review. We excluded 1089 records based on title and abstract
seizure outcome data were recorded) and 383 met other exclusion and assessed 270 full-text articles for inclusion in the review. We
criteria of the review (see Tonini 2004 for further details). Forty- excluded 93 studies from the review (see Excluded studies below)
seven studies were included in the Tonini 2004 review; we screened and included 177 studies in the review (see Included studies below).
these studies for inclusion in our review as 'records found from See Figure 1 for PRISMA study flow diagram
other sources'.


Informed decisions.

Figure 1. Study flow diagram.

Informed decisions.

Included studies participants at surgery (see Table 1) was available for 152 studies
(86% of total studies), ranging from 0 years to 86 years at surgery.
We included 177 studies in this review (Aaberg 2012; Adam 1996;
Age at surgery was not available for 2707 participants from 25
Adelson 1992; Alfstad 2011; Althausen 2013; Arruda 1996; Awad
studies (14% of total studies). Assuming that adults are classified
1991; Babini 2013; Battaglia 2006; Baumann 2007; Bautista 2003;
as over the age of 18 years, 28 studies included adults only (16% of
Bell 2009; Benifla 2006; Berkovic 1995; Blount 2004; Blume 2004;
total studies, 2285 participants), 21 studies included children only
Boesebeck 2007; Boshuisen 2010; Brainer-Lima 1996; Britton 1994;
(12% of total studies, 1147 participants) and 103 studies included
Caraballo 2011; Cascino 1995; Chabardes 2005; Chang 2009; Chee
both adults and children (58% of total studies, 10,374 participants).
1993; Chkhenkeli 2007; Choi 2004a; Chung 2005; Cossu 2005; Cossu
A measure of the duration of epilepsy of the participants (see
2008; Costello 2009; Cukiert 2002; Dagar 2011; Dalmagro 2005;
Table 1) was available for 110 studies (62% of total studies, 10,391
Delbeke 1996; Dellabadia 2002; de Tisi 2011; Devlin 2003; Donadio
participants) ranging from 0 years to 86 years. Duration of epilepsy
2011; Dorward 2011; Duchowny 1998; Dunkley 2011; Dunlea 2010;
was not available for 6117 participants from 67 studies (38% of total
Elsharkawy 2008a; Elsharkawy 2009a; Elsharkawy 2011a; Engman
studies).
2004; Erba 1992; Erickson 2005; Fauser 2004; Fujiwara 2012; Garcia
1991; Garcia 1994; Gelinas 2011; Georgakoulias 2008; Gilliam The type of surgical resection performed was available for 168
1997a; Gilliam 1997b; Goldstein 1996; Greiner 2011; Grivas 2006; studies (95% of total studies). In 91 studies a single-lobe surgery
Gyimesi 2007; Hader 2004; Hajek 2009; Hallbook 2010; Hamiwka was reported; 7942 participants from 76 studies (43% of total
2005; Hartley 2002; Hartzfield 2008; Hemb 2010; Holmes 1997; studies) underwent temporal lobe resection only, 1058 participants
Holmes 2000; Jack 1992; Janszky 2003a; Janszky 2003b; Jaramillo- from 15 studies (8.5% of total studies) underwent extratemporal
Betancur 2009; Jayakar 2008; Jayalakshmi 2011; Jeha 2006; Jehi lobe resection only and in 77 studies (43.5% of total studies)
2012; Jennum 1993; Jeong 1999; Kan 2008; Kang 2009; Kanner including 6661 participants both temporal and extratemporal lobe
2009; Kilpatrick 1997; Kim 2009; Kim 2010a; Kim 2010b; Kloss 2002; resection were reported.
Knowlton 2008; Kral 2007; Krsek 2013; Kuzniecky 1993; Kwan 2010;
Lackmayer 2013; Lee 2006; Lee 2008; Lee 2010a; Lee 2011; Lei One hundred and forty-four studies identified 13,557 participants
2008; Li 1997; Li 1999; Liang 2010; Liang 2012; Liava 2012; Lopez- via a retrospective design (81% of 177 studies), 21 studies identified
Gonzalez 2012; Lorenzo 1995; Madhavan 2007; Mani 2006; Mathern 1872 participants via a prospective design (12% of studies), three
1999; McIntosh 2012; Mihara 2004; Miserocchi 2013; Morino 2009; studies identified 342 participants via a combination of prospective
Morris 1998; O'Brien 1996; O'Brien 2000; Oertel 2005; Paglioli 2006; and retrospective designs (2% of studies) and in nine studies the
Paolicchi 2000; Park 2002; Park 2006; Perego 2009; Perry 2010; method of identification of the 670 participants was not stated and
Phi 2009; Phi 2010; Pinheiro-Martins 2012; Prevedello 2000; Raabe could not be deduced (5% of studies).
2012; Radhakrishnan 1998; Rausch 2003; Remi 2011; Roberti 2007;
Rossi 1994; Russo 2003; Sagher 2012; Sakamoto 2009; Salanova Follow-up in the 177 studies ranged from 0 to 366 months. We
1994; Sarkis 2012; Schramm 2011; Seymour 2012; Sinclair 2003; specified in our inclusion criteria (see Criteria for considering
Sindou 2006; Sola 2005; Spencer 2005; Sperling 1992; Stavrou 2008; studies for this review) that studies must have an 'Expected
Suppiah 2009; Swartz 1992; Tanriverdi 2010; Tatum 2008; Terra- duration of follow-up of at least one year' for inclusion, therefore
Bustamante 2005a; Terra-Bustamante 2005b; Tezer 2008; Theodore we excluded studies in which the duration of follow-up was not
2012; Tigaran 2003; Tripathi 2008; Trottier 2008; Urbach 2007; specified at all, the duration of follow-up was definitely less than
Ure 2009; Velasco 2011; Walz 2003; Weinand 1992; Wellmer 2012; one year or if it was unclear how many participants were followed
Widdess-Walsh 2007; Wiebe 2001; Wieshmann 2008; Wray 2012; up for at least one year. A number of studies included participants
Wyler 1995; Wyllie 1998; Yang 2011; Yeon 2009; Yu 2009; Yu 2012a; with less than 12 months follow-up but did not include these
Yu 2012b; Zangaladze 2008; Zentner 1995; Zentner 1996). participants in postoperative seizure outcomes or it was possible
for us to separate data for participants followed up for less than
We included four studies of a randomised controlled design. Three or more than 12 months. We considered these studies to have an
studies randomised the type or length of surgical intervention: expected follow-up of more than one year. Similarly, we considered
anterior temporal lobectomy with or without anterior corpus studies in which a mean or median follow-up time greater than 12
callosotomy (Liang 2010), partial versus total hippocampectomy months was reported but no minimum follow-up was reported to
(Wyler 1995), and 2.5 cm or 3.5 cm tailored temporal lobe resection have an expected duration of follow-up of at least 12 months. At
(Schramm 2011). One study randomised participants to immediate least two review authors closely examined follow-up information
surgery (n = 40) or medical treatment (antiepileptic drugs, n = 40) in all studies before a decision was made to include or exclude the
and placed on a waiting list for surgery (Wiebe 2001). study.
All other studies were of a non-randomised design and did not We defined three types of outcome scale appropriate for the
include a control group in the study design. primary seizure outcome (see Primary outcomes). Forty-two
studies reported seizure outcome according to 'more than one year
See Characteristics of included studies and Table 1 for detailed seizure-free scale' (24% of total studies, 3981 participants), 117
study characteristics and participant demographics in all 177 studies reported seizure outcome according to the 'Engel Class
included studies. In summary: Scale' (65% of total studies, 10,619 participants) and 18 studies
reported seizure outcome according to an 'other scale' (11% of total
A total of 16,508 participants in the 177 studies underwent a surgical
studies, 1908 participants). Other scales we deemed to meet our
procedure for intractable epileptic seizures. Gender was reported
definition of a 'good outcome' were as follows:
for 13,360 participants from 149 studies (7591 male (57%) and
5769 female (43%)) and was missing for 3148 participants from 28 Seven studies reported more than two years seizure freedom
studies (19% of total participants). A measure of the age of the (Engman 2004; Goldstein 1996; Gyimesi 2007; Holmes 2000; Jeong
Informed decisions.

1999; Lee 2006; Mathern 1999), one reported more than three years Placantonakis 2010; Rocamora 2009; Sakuta 2005; Soeder 2009;
seizure freedom (Rossi 1994), and one reported more than five years Teutonico 2008; Upchurch 2010; Wetjen 2009; Yasuda 2010b). We
seizure freedom (McIntosh 2012). excluded 20 studies that did not report any seizure outcome
data (Alemany-Rosales 2011; Andersson-Roswall 2010; Baxendale
Five studies reported seizure outcome according to the 2005; Bell 2010; Carne 2004; D'Argenzio 2011; Elsharkawy 2009b;
classification proposed by the International League Against Fauser 2008; Griffin 2007; Hervas-Navidad 2002; Hildebrandt 2005;
Epilepsy (ILAE 2001) as follows: Lutz 2004; McClelland 2007; McClelland 2011; Oertel 2004; Park
2010; Stavem 2005; Stavem 2008; Wetjen 2006; Yasuda 2010a).
Class 1, completely seizure-free; Class 2, aura alone with no seizure;
We excluded eight studies in which the inclusion criteria were
Class 3, one to three seizure days/year; Class 4, 50% reduction in
based on seizure outcome (i.e. only participants with postoperative
baseline number of seizure days; Class 5, less than 50% reduction
seizures or who were seizure-free were included) (Boshuisen
in baseline number of seizure days and Class 6, more than 100%
2012; Buckingham 2010; Elsharkawy 2011b; Jehi 2010; Lach 2010;
increase in baseline number of seizure days.
Schwartz 2006; Stefan 2008; Vadlamudi 2004). We excluded eight
We considered ILAE Class 1 to correspond to Engel Class 1A and duplicate studies and retained the primary reference in the review
ILAE Classes 1 and 2 together to correspond to Engel Class 1 (see (Boesebeck 2002; Cascino 1996; Elsharkawy 2008b; Helmstaedter
Primary outcomes), therefore if a study reported a 'good outcome' 2011; Kuzniecky 1996; Lachhwani 2003; Malla 1998; Weinand 2001).
to be ILAE Class 1 (poor outcome classes 2 to 6) or ILAE Classes 1 and We found no papers from the same institutions reporting different
2 (poor outcome classes 3 to 6), we accepted this as a satisfactory factors (all reported either the same factors or factors outside the
'other scale'. Three studies defined ILAE Class 1 as a good outcome scope of our review). Characteristics of excluded studies carries
(de Tisi 2011; Kral 2007; Lackmayer 2013), and two defined ILAE the relevant detail. We excluded three studies in which fewer than
classes 1 and 2 as a good outcome (Sakamoto 2009; Yang 2011). 90% of participants had an MRI (Mohammed 2012; Wieser 2003a;
We did not accept any other combinations of ILAE classes as an Wieser 2003b), and we excluded two studies that were not studies
outcome that measured seizure freedom of surgery for epilepsy (one was a study of surgery for tumours
(Grunert 2003), and one was a study of the outcome of taking
Three studies reported seizure freedom by the Engel Class Scale but antiepileptic drugs after epilepsy surgery (Asadi-Pooya 2008)).
not as our definition above, where a good outcome corresponds
to Engel Class 1 and a poor outcome corresponds to Engel Class 2 Risk of bias in included studies
to 4 (see Primary outcomes); two studies defined a good outcome As the majority of studies in this review were not of a randomised
as Engel Class 1A and a poor outcome as Engel Classes 1B to 4 design, we did not feel it was appropriate to judge the quality of
(Boshuisen 2010; Phi 2010), and one study defined a good outcome each study on criteria of selection bias, performance and detection
as Engel Class 1A to B and a poor outcome as Engel Classes 1C bias, attrition bias and reporting bias. Instead, we employed the
to 4 (Boesebeck 2007). As the definition in these studies did not Effective Public Health Practice Project (EPHPP) tool, which is
match our definition of the Engel Class Scale but these definitions appropriate for study designs included in this review (mostly
do measure seizure freedom, we deemed these three studies to retrospective case series). See Appendix 4.for the full tool, Table
report a satisfactory 'other scale'. 2 for quality assessments for each study for each criterion of the
tool (A to F) and an overall grading of quality for each study and
Four out of the 177 included studies (255 participants) did not
Table 3 for a summary of all components of the tool (A to G).
report an outcome scale that we deemed satisfactory for measuring
We note that for the purposes of analysis Yu 2012a and Yu 2012b
seizure freedom; three studies defined a good outcome as Engel
were treated as separate studies within a single publication (adults
Classes 1 and 2 and a poor outcome as Engel Classes 3 to 4(Kwan
and children respectively), however for the purposes of quality
2010; Krsek 2013; Ure 2009), and one study defined a good outcome
assessment, we considered the publication only once therefore
as 'seizure free or rare seizures' and a poor outcome as 'less than
quality assessments were made for 176 studies in this review. We
80% reduction in seizures' (Kuzniecky 1993). These four studies
also note that the EPHPP tool is a tool for quality assessment rather
were not included in meta-analysis (see Effects of interventions).
than risk of bias so we refer to 'quality assessment of included
Excluded studies studies' throughout this section.
We excluded a total of 93 studies from the review. We excluded A. Selection bias

30 studies with insufficient follow-up (in other words follow-up
We judged that 124 out of 176 studies (70%) recruited a sample of
was not defined, follow-up was less than a year or an unknown
individuals that were 'very likely' to be representative of the target
proportion of participants were followed up for less than a
population. We made this judgement if the participants recruited
year (Acar 2008; Alpherts 2008; Binder 2009; Busch 2011; Chang
were 'consecutive' or if all eligible participants undergoing surgery
2007; Choi 2004b; Cohen-Gadol 2003; Colonnelli 2012; Coutin-
over a specific period of time were included. We also checked
Churchman 2012; D'Angelo 2006; da Costa-Neves 2012; Dulay
how reasonable inclusion criteria, exclusion criteria and participant
2006; Dulay 2009; Ferrari-Marinho 2012; Ferroli 2006; Freitag 2005;
demographics in the study were for recruiting a sample very likely
Ghacibeh 2009; Harvey 2008; Hellwig 2012; Helmstaedter 2004; Hu
to be representative of the target population.
2012; Junna 2013; Limbrick 2009; Ogiwara 2010; Roth 2011; Smyth
2007; Stefan 2004; Vachrajani 2012; Vadera 2012; Zupanc 2010). We judged that the remaining 52 studies (30%) recruited samples
We excluded 22 studies with fewer than 30 participants (Bauer of individuals that were 'somewhat likely' to be representative
2007; Bourgeois 2007; Caicoya 2007; Cukiert 2009; Danielsson 2009; of the target population. We made this judgement if participants
Datta 2009; Engel Jr 2012; Haegelen 2013; Lee 2010; Lodenkemper were not consecutive or if apparently eligible participants had been
2007; Mikati 2004; Moien-Afshari 2009; Negishi 2011; Nikase 2007; excluded. We also made this judgement if we were uncertain over

Informed decisions.

where participants had been recruited from, if we were uncertain randomised by a computer-generated block randomisation list.
regarding inclusion or exclusion criteria, if we judged that inclusion Two studies were described as randomised but no details were
and exclusion criteria were too specific or restrictive to recruit a given about the method of generation of the random list (Wiebe
sample representative of the target population (e.g. all participants 2001; Wyler 1995). One study described an inadequate method of
had a very specific pathology for inclusion) or if we judged that the quasi-randomisation, which we judged to be high risk of bias (Liang
demographics of recruited participants were not representative of 2010). Two studies did not provide any information on concealment
the target population. of treatment allocation (Liang 2010; Wyler 1995), and two studies
described adequate methods of allocation concealment: sealed,
We judged that if the sample recruited was based in any way on opaque, sequentially numbered envelopes prepared outside of
outcome (for example, only individuals with recurrence of seizures the treatment centre (Schramm 2011; Wiebe 2001). One study did
included), then the study had recruited a sample of individuals not provide any information on blinding of participants, personnel
'not likely' to be representative of the target population. Such a and outcome assessors (Liang 2010), and in one study in which
study design met the exclusion criteria for this review (see Excluded participants were randomised to surgical or medical treatment
studies). Therefore we judged all included studies in the review to blinding was not possible by design. However, it is unclear if
have a sample of individuals 'very likely' or 'somewhat likely' to be outcomes were affected by this design (Wiebe 2001). Two studies
representative of the target population. reported blinding of participants and outcome assessors with
only the surgeon remaining unblinded to allocation of treatment
In the 143 studies of a retrospective design we were unable to judge
(Schramm 2011; Wyler 1995). In three studies no losses to follow-
the percentage of participants who agreed to take part in the study.
up occurred and all randomised participants were included in the
In six studies of a prospective design the percentage of participants
analysis (Liang 2010; Schramm 2011; Wyler 1995). In one study
who agreed to take part in the study was reported and was 80% to
complete attrition rates were reported and an intention-to-treat
100% for all six studies. In the remaining 27 studies, we could not
approach was taken (Wiebe 2001), so we judged all four studies
tell how many participants agreed to take part in the study.
to be at low risk of attrition bias. We did not have access to study
Overall we judged that 114 studies (65%) were of 'strong' quality in protocols to compare to outcomes defined a priori but all studies
their selection criteria; i.e. the sample selected were very likely to reported outcomes defined in the methods well and consistently in
be representative of the target population and the study was of a the results section, so we judged all four studies to be at low risk of
retrospective design or 80% to 100% of participants agreed to take reporting bias. We detected no other biases in any of the studies.
part for studies of a prospective design. We judged the remaining 62
Two studies described a randomised design: Oertel 2005
studies (35%) to be of 'moderate' quality in their selection criteria;
randomised participants to waterjet dissection or ultrasonic
i.e. the sample selected were somewhat likely to be representative
aspirator during surgery and Velasco 2011 randomised participants
of the target population or the sample selected were very likely to
to presurgical evaluation with or without single-photon emission
be representative of the target population but we could not tell how
computed tomography (SPECT). As neither of these designs
many participants had agreed to participate in the study.
randomised the intervention, for the purposes of this review, we
B. Study design refer to these designs as 'cohort analytic' (two groups pre and post
intervention) and we also considered this design to be of 'strong'
The majority of the studies included (143 studies out of 176 studies, quality according to the EPHPP tool.
81%) were of the design of a retrospective review of the clinical
notes of a number of participants meeting specific inclusion criteria The remaining 27 studies (15 of a prospective design, three of a
at a given centre over a specified time period (for example, a clinical combination design and nine with the method of identification of
audit). We refer to this design in Table 2, Table 3 and throughout participants not stated), we considered to be of a 'cohort' design
this review as a 'retrospective case series' (one group pre and post (one group pre and post intervention) and this design is considered
intervention) and we considered this design to be of 'moderate' to be of 'moderate' quality according to the tool.
quality as this design was not specifically referred to in the EPHPP
tool. We made this judgement to avoid the design of the study alone Therefore, we judged six out of 176 studies (3%) to be of a 'strong'
dictating the overall quality rating (i.e. all retrospective designs are design and 170 out of 176 to be of a 'moderate' design.
weak) in order to separate out the 143 retrospective studies based
C. Confounders
on other quality criteria.
Given that the aim of our review was to identify prognostic
We included four studies of a randomised controlled design, a factors associated with the outcome of surgery (essentially factors
design judged to be strong by the EPHPP tool (Liang 2010; Schramm that confound the results of surgery), it was difficult to make a
2011; Wiebe 2001; Wyler 1995; see Included studies for further judgement on the presence of confounders. Therefore we did not
details). We also used the Cochrane 'Risk of bias' tool for these class any of the pre-operative prognostic factors of interest in our
four studies (Higgins 2011), see Table 4 for more information on the review (see Data extraction and management) as confounders.
'Risk of bias' criteria for these four studies. We preferred to present Furthermore, the majority of studies (170 out of 176 studies, 97%)
quality assessment data in Additional tables rather than in the 'Risk were of a design which followed up one group of participants pre
of bias' tables in Characteristics of included studies as the domains and post intervention (retrospectively or prospectively) rather than
considered in these tables are not appropriate for the majority of two groups that may differ in terms of demographics, so by this
the included studies. single sample design, it is more appropriate to judge the methods
of selection of the sample (see A. Selection bias) than to assess
Table 4 shows that only one of the four RCTs described an adequate
'confounders'. For this reason, we judged these 170 studies to be of
method of randomisation; participants in Schramm 2011 were
'strong' quality, in the absence of confounding variables.

Informed decisions.

In the four studies of a randomised controlled design (Liang E. Data collection methods
2010; Schramm 2011; Wiebe 2001; Wyler 1995), and the two
Our outcome of interest in this review was seizure freedom
studies of a cohort analytic design (Oertel 2005; Velasco 2011),
following surgery, therefore the data required for the outcome
with two groups pre and post intervention (randomised and
would be details of recurrence of participant seizures after surgery.
non-randomised respectively), we made a judgement regarding
Given that such data are recorded from participant reports, which
potential confounding factors (other than prognostic factors) in
may be prone to recall error and there are no validated tools (such
the two groups. Five of these studies presented demographics
as quality of life assessment tools etc.) to record such data, for each
for the two participant groups and/or tested whether any
study given the information reported on outcome data collection,
significant differences were present between the groups (Liang
we made a judgement on whether the methods were adequate and
2010; Schramm 2011; Velasco 2011; Wiebe 2001; Wyler 1995). No
reliable.
significant differences were found in any of these studies therefore
we judged that groups were balanced and there was no evidence of Overall, we accepted any method of seizure data collection that
confounders. In one study, very limited information was reported seemed reasonable (participant seizure diaries, clinical notes,
on the demographics of the two groups and it did not appear that interviews with participants and/or family members at clinic visits
the groups had been compared for differences which may have or over the phone, postoperative MRI or EEG) to be valid. We judged
influenced results of analysis (Oertel 2005); for this study we could all of these methods to be 'reliable' given that all methods are
not tell if any confounders were present or if any adjustments had likely to be prone to error as the outcome is somewhat subjective.
been made to the analysis to account for confounders. Ninety out of 176 studies (51%) reported a 'valid and reliable' data
collection method and we judged these studies to provide 'strong'
Therefore we judged that for 175 studies (99%) the quality of
quality evidence.
evidence was strong; i.e. no confounders that may have influenced
the results of analysis were present. For one study of a cohort The remaining 86 studies (49%) either did not provide any
analytic design, the quality of evidence was moderate; i.e. we were information at all on data collection methods or did not provide
unable to make a judgement whether any confounders which may sufficiently clear information on data collection methods for us to
have influenced analysis were present (Oertel 2005). judge them to be valid and/or reliable. We judged these studies to
provide 'weak' quality evidence.
D. Blinding
In the context of surgical treatment, blinding is very difficult F. Withdrawals and drop-outs
as the operating surgeon is required to know the procedure This criterion was not applicable for the 143 studies (81%) of a
being carried out and the participant will be aware that he/ retrospective design. For the remaining 33 studies we made a
she is undergoing surgery. It may be possible, however, to blind judgement on whether withdrawal information was adequately
participants and outcome assessors (other than the surgeon) to reported.
the specific surgical procedure being carried out and outcome
assessors were often blinded to pre-operative evaluation details In 12 studies, all of a prospective design, the number of
when making judgements regarding outcome following surgery. We withdrawals/losses to follow-up in the study were reported with
believe that in the context of surgical treatment, where blinding is reasons where applicable; in all of these 12 studies at least 80%
usually unfeasible, the outcomes of the study would not necessarily of participants completed the study; less than 20% withdrew from
be influenced. However, for the purposes of this quality assessment the study. We judged these studies to provide a 'strong' quality
we followed the criteria specified by the EPHPP tool relating to evidence (7% of total studies). In the remaining 23 studies (nine of
blinding of the intervention. a prospective design, three of a combination design and nine with
design not stated), no information was given regarding participant
Question 2 of the EPHPP tool for the blinding component (see withdrawal or losses to follow-up therefore we could not tell how
Appendix 4) is as follows: 'Were the study participants aware of the many participants completed the study. We judged these studies to
research question?' We interpreted this to mean 'Were the study provide 'weak' quality evidence (21% of total studies).
participants aware of the intervention allocated?' as we believe
that awareness of the research question is not the same thing as G. Intervention integrity
blinding to allocation of intervention and it would be impossible to
determine the awareness of participants of a research question in In the context of a surgical intervention, we judged that it was highly
a study of retrospective design. unlikely that any patient received an unintended intervention.
Within all 176 included studies, 80% to 100% of participants
In 173 out of 176 studies (98%) we judged that outcome assessors included in the study received the surgical intervention. Several
and participants were aware of the intervention and that these studies (particularly those aiming to identify criteria in presurgical
studied provided 'weak' quality evidence due to high risk of evaluation that may be associated with outcome) specified
bias from lack of blinding. The remaining three studies were numbers of participants who underwent presurgical evaluation
randomised controlled trials. Two studies blinded both participants and were not recommended for surgical intervention, however this
and outcome assessors and we considered these studies to be proportion of participants was less than 20% for all studies and only
'strong' quality evidence (Schramm 2011; Wyler 1995). One trial did participants who underwent surgery were included in the results of
not provide any information on blinding of participants or outcome this review.
assessors so we judged this to be of 'moderate' quality evidence
(Liang 2010). We judged that the 'consistency' of the intervention had been
measured if a study reported at least details of all surgical
techniques used for all participants; it was not necessary for the

Informed decisions.

surgical intervention to be exactly the same for all participants and retrospective, would be the most appropriate tool to use for
in a study for the intervention to be considered consistent as this review.
many participants required tailored resections based on pathology
or aetiology of seizures. We also considered the intervention Following use of the tool by five authors for the 176 included
to be 'consistent' if the same surgical technique was used for studies, we now feel that this tool has not provided a fully accurate
all participants, the same surgical protocol was used, the same assessment of the quality of the included studies. Assuming a
surgeon(s) performed all surgeries and if any specific differences in randomised controlled trial to be the highest quality evidence of
surgical technique were reported. We judged that the 'consistency' the efficacy of an intervention (Mann 1996), 81% of the studies
of the intervention had not been measured if the details of surgical included in this review were of a retrospective, single-group design,
techniques were not reported or not reported for all participants which could be considered as the poorest quality in the hierarchy
or if the type of intervention(s) performed were unclear. We judged of evidence about an intervention. Therefore we needed a tool to
that the intervention was consistent for 158 studies (90%) and separate the different levels of 'poor' evidence and the EPHPP tool
inconsistent for 18 studies (10%). did not do this.
H. Analyses Despite the described applicability of this tool to all quantitative

study designs, many of the criteria were not appropriate for
All studies had individual units of allocation and analysis by design; studies of a retrospective design (e.g. withdrawals from the study,
in other words in all studies a surgical intervention was performed proportion of participants agreeing to take part in the study) and
on each individual and each individual was analysed for the seizure studies of a single-group design (confounders). Also in the context
outcome. We identified no studies of a cluster (randomised) design of surgical studies for epilepsy where blinding of participants and
for this review. We judged that all studies performed analysis outcome assessors is often impossible and only objective, non-
by intervention allocation status (intention-to-treat) rather than validated methods for the collection of seizure outcome data exist,
actual intervention received; in studies of a retrospective design the global quality rating of the study was influenced by two of
it was difficult to know if a different type of surgical intervention the components (D and E) for most of the studies. Blinding was
had been 'allocated' as the only information available from a attempted in only three randomised controlled trials included in
retrospective study is the intervention received. For studies of a this review, therefore the judgement for 173 out of 176 studies
prospective or combination design we judged that all studies had (98%) was automatically assigned to be 'weak' according to the
taken an intention-to-treat approach to analysis. definition of the tool. On the basis of global rating, this meant that
98% could be judged at the most to be of 'moderate' quality due to
One hundred and fifty-three out of 176 studies (87%) performed
lack of blinding alone. Then the difference between a 'moderate' or
statistical analysis and the remaining 23 studies (13%) reported
'weak' global rating was dictated by component E (strong or weak).
only observational results without performing analysis. The
Essentially the global rating reflects the quality of data collection
statistical author of this review (SN) judged all statistical analyses
methods rather than overall quality.
performed in the 153 studies to be appropriate for the study design.
Furthermore, for component B. Study design, we made an
Overall rating
assumption that the 143 'retrospective case series' (one group pre
The global quality rating was based on components A to F of the and post intervention), not specifically referred to in the EPHPP
EPHPP tool. We judged a study that had no components judged tool, were of a 'moderate' quality study design; the same quality
as 'weak' as 'strong' overall, one 'weak' component as 'moderate' assigned to a prospective cohort study. We made this judgement
overall and two or more 'weak' components as 'weak' overall. to avoid the design of the study alone dictating the overall quality
rating (i.e. all retrospective designs are weak) in order to separate
We judged the global quality rating to be 'strong' for two studies out the 143 studies based on other quality criteria. If instead,
(1% of total studies, both randomised controlled trials, Liang 2010 more fitting with NICE guidelines (Mann 1996), we had assumed all
and Wyler 1995), 78 studies to be of 'moderate' quality (44% of total retrospective designs to be of 'weak' quality then the global rating
studies) and 96 studies to be of 'weak' quality (55% of total studies). would have been as follows: strong: two studies; moderate: four
studies; weak: 170 studies. Such ratings would not have provided
Adequacy of EPHPP quality assessment tool us with any useful information regarding the relative quality of
Following on from the work of Tonini 2004 related to this review, included studies.
we knew that the majority of the studies identified via searches for
this review were likely to be of a non-randomised and retrospective In hindsight, given the context of surgery for epilepsy and our prior
design. At the initiation of the protocol for this review in 2012, knowledge of the likely design of included studies based on the
a 'Risk of bias' tool for the assessment of randomised controlled Tonini 2004 review, it would have been more appropriate for us
trials had been developed and recommended by the Cochrane to design our own quality assessment tool for the review based
Handbook for Systematic Reviews of Interventions (chapter 8, on what we know to be clinically important in studies of surgical
Higgins 2011), however we were not aware of a specific tool for the interventions. Given our lack of confidence in the global ratings
assessment of studies of a non-randomised and/or retrospective assigned to the included studies by the EPHPP tool, we felt it
design recommended by the Cochrane Handbook for Systematic would be inappropriate to conduct sensitivity analyses based on
Reviews of Interventions (chapter 13, Higgins 2011). Therefore two the global quality assessments (see Effects of interventions for
authors (SN and JP) with methodological experience of quality subgroup and sensitivity analyses performed).
assessment reviewed the existing literature at the time for a quality
assessment tool and judged that the EPHPP tool, intended for the
quality assessment of all study designs including non-randomised

Informed decisions.

Assessment of risk of bias in studies reporting multivariable 2012; Sagher 2012; Sarkis 2012). We judged two studies to be at
prognostic models according to the Quality in Prognostic high risk of measurement bias; in one multi-centre study it is likely
Studies (QUIPS) tool that data were collected using different methods across centres
(Madhavan 2007), and in one study a large proportion of prognostic
Twenty-eight studies reported a multivariable prognostic model
factor data was missing, which is likely to have had an impact on
including one or more of the factors of interest to us (Althausen
analyses (Rossi 1994).
2013; Boesebeck 2007; Cossu 2005; Cossu 2008; Elsharkawy 2008a;
Elsharkawy 2009a; Gelinas 2011; Janszky 2003a; Jennum 1993; Kim 4. Outcome measurement: judge the risk of bias related to the
2009; Kim 2010a; Lopez-Gonzalez 2012; Madhavan 2007; McIntosh measurement of outcome (differential measurement of outcome
2012; O'Brien 2000; Paolicchi 2000; Phi 2009; Radhakrishnan 1998; related to the baseline level of PF)
Rossi 1994; Sagher 2012; Sarkis 2012; Schramm 2011; Spencer 2005;
Tezer 2008; Theodore 2012; Walz 2003; Wyler 1995; Yang 2011), see Recurrence of seizures in outpatients is generally patient-reported
'Multivariable analyses' in Effects of interventions below for more and therefore difficult to measure in a valid and reliable way;
details of these results and results reported. validated scales such as the Engel Class Scale as described above
exist for assessment of post-surgical outcome.
Six domains are considered in the QUIPS tool (see Table 5 for
judgements for each domain for each study): We judged 18 out of 28 studies to be at moderate risk of bias where
it was unclear exactly how and/or when outcome data had been
1. Study participation: judge the risk of selection bias (likelihood collected and/or how outcome was defined. We judged two studies
that the relationship between Prognostic Factor (PF) and to be at high risk of bias; in one study, Althausen 2013, outcome
outcome is different for participants and eligible non- was measured in variable ways (taken only from patient reports
participants) for some participants and using supplementary data from medical
records for some participants) and in one study, Kim 2009, the
Eleven out of 28 studies are judged to be at low risk of selection
outcome was not measured according to a known scale such as
bias; the population of interest and the method of sample
the Engel Class Scale and no information was available on when
recruitment are well described and the sample seem to match the
outcome was recorded. We judged the remaining eight studies to be
characteristics of the source population. We judged 15 of the 28
at low risk of bias where clear information was reported about how
studies to be at moderate risk of selection bias due to uncertainties
outcome data were collected, outcome was measured according to
or limited information regarding the population of interest or the
a known scale such as the Engel Class Scale and was recorded at
method of sample recruitment, or both, to be able to judge whether
the same time for all participants.
the sample matched the source population. We judged two studies
to be at high risk of selection bias: one due to a selective sample, 5. Study confounding: judge the risk of bias due to confounding
which is unlikely to represent the source population (Janszky (i.e. the effect of PF is distorted by another factor that is related
2003a), and one due to very limited information regarding the to PF and outcome)
population of interest, the method of sample recruitment and the
characteristics of the sample (Rossi 1994). Twenty of the 28 studies were of a single-group design so this
domain was not applicable. We judged one RCT, Wyler 1995, to be
2. Study attrition: judge the risk of attrition bias (likelihood at low risk of bias due to confounding as the randomised design
that the relationship between PF and outcome are different for should remove confounding and it is confirmed in the study that
completing and non-completing participants) groups were balanced at baseline. We judged seven studies to be at
moderate risk of bias of confounding: one study of an RCT design
Twenty-two of the 28 studies were of a retrospective design so this
did not demonstrate whether groups were balanced at baseline
domain was not applicable. For the six studies of a prospective
and which variables of interest were prognostic factors and which
design, in two studies we judged attrition bias to be at low risk
were confounders (Schramm 2011), one study makes reference to
of attrition bias as intention-to-treat analyses were planned in
confounders and interactions in a generalised estimating equations
the case of withdrawals or losses to follow-up, so all participants
model but does not specify which variables of interest were
contributed to outcome assessment (Schramm 2011; Wyler 1995).
prognostic factors and which were confounders (Sagher 2012),
We judged three to be at moderate risk of attrition bias due to lack
and the remaining five studies make reference to "confounders"
of information reported about withdrawals and losses to follow-
but do not adequately define the variables and/or do not specify
up (Spencer 2005; Theodore 2012; Walz 2003), and one study to
which variables of interest were prognostic factors and which
be at high risk of attrition bias due to exclusion of participants
were confounders (Althausen 2013; Cossu 2008; Elsharkawy 2008a;
with missing data and uncertainty over whether participants were
Gelinas 2011; Janszky 2003a).
recruited prospectively or retrospectively (Radhakrishnan 1998).
6. Statistical analysis and reporting: judge the risk of bias
3. Prognostic factor measurement: judge the risk of
related to the statistical analysis and presentation of results
measurement bias related to how PF was measured (differential
measurement of PF related to the level of outcome) We judged only one study to be at low risk of bias (O'Brien
2000), where statistical analysis was well described, modelling
We judged 22 out of the 28 studies to be at moderate risk
was performed based on clinical relevance and results were not
of measurement bias due to unclear definitions of prognostic
selectively reported. We judged the other 27 studies to be at
factors and limited information regarding how data were collected.
moderate or high risk of bias (13 moderate and 14 high) due to
Detailed definitions of prognostic factors and methods of
unclear or inappropriate statistical methods, selection of variables
measurement and data collection were reported in four studies
based on statistical significance and selective reporting of results.
judged to be at low risk of measurement bias (Cossu 2008; McIntosh
These issues and the likely impact on the analyses are discussed

Informed decisions.

further in 'Multivariable analyses' in Effects of interventions Secondary outcome: Seizure outcome according to prognostic
below. factors of interest
Overall, we judged one study to be at high risk of bias in three Wiebe 2001 did not report any univariate or multivariable analyses
domains (Rossi 1994), three studies to be at high risk of bias in that investigated the influence of any prognostic factors on the
two domains(Althausen 2013; Janszky 2003a; Madhavan 2007), 11 seizure outcome (including the prognostic factors of interest to us).
studies to be at high risk of bias in one domain (Cossu 2008;
Overall outcome according to surgical techniques
Elsharkawy 2008a; Elsharkawy 2009a; Jennum 1993; Kim 2009;
Paolicchi 2000; Radhakrishnan 1998; Sarkis 2012; Spencer 2005; One study randomised participants to anterior temporal lobectomy
Tezer 2008; Theodore 2012) and we did not judge 13 studies to be with or without anterior corpus callosotomy (ATL versus aCCT)
at high risk of bias in any domain (Boesebeck 2007; Cossu 2005; (Liang 2010). Thirty participants were randomised to each group
Gelinas 2011; Kim 2010a; Lopez-Gonzalez 2012; McIntosh 2012; and followed up for two years for assessment of outcome.
O'Brien 2000; Phi 2009;Sagher 2012; Schramm 2011; Walz 2003;
Wyler 1995; Yang 2011). Transient complications of surgery were reported in nine
participants in total. In the aCCT group there were two cases of
Effects of interventions urinary incontinence, one case of aphasia and two cases of apraxia.
In the ATL group there were two cases of aphasia and two cases of
See: Summary of findings for the main comparison; Summary of apraxia.
findings 2
One study randomised participants to a 2.5 cm or 3.5 cm
Overall outcome of surgery compared to medical treatment tailored temporal lobe resection (i.e. an intended minimum
One study randomised participants to a surgical or control resection length of 25 versus 35 mm for the hippocampus
intervention (Wiebe 2001). Forty participants were randomised and parahippocampus) (Schramm 2011). One hundred and four
to each intervention group and followed up for 12 months for participants were randomised to a 2.5 cm resection and 103 were
assessment of outcome. randomised to a 3.5cm resection and were followed up for 12
months for assessment of outcome.
Four participants allocated to the surgery group did not undergo
surgical intervention (one declined surgery, two were deemed There were no significant differences between the 2.5 cm and
not eligible for surgery based on pre-operative testing and one 3.5 cm resection groups concerning neurological complications (P
did not have seizures during pre-operative testing). An intention- value < 0.605), visual field defects (P value < 0.856) or surgical
to-treat approach was taken to analysis and all randomised complications (P value < 0.875). Details of surgical complications
participants were analysed in the allocated groups, regardless of were not given.
the intervention received.
One study randomised participants to partial (hippocampus was
Four participants experienced adverse events from surgery: one removed en bloc to either the anterior margin of the cerebral
had a small thalamic infarct causing sensory abnormalities peduncle) or total (hippocampus was removed en bloc to the level
in the thigh, one had a wound infection, two had decline of the colliculi) hippocampectomy and followed up participants
in verbal memory that interfered with their occupations at for 12 months for assessment of outcome (Wyler 1995). Thirty-four
one year. In addition, 22 participants in the surgery group participants were randomised to partial hippocampectomy and 36
experienced asymptomatic, superior sub-quadrantic visual field to total hippocampectomy.
defects, seven participants experienced depression and one
participant developed transient psychosis. The only adverse event There were complications in five participants: two participants
reported in the medical treatment group was depression (eight with partial hippocampectomy (one subgaleal cerebrospinal fluid
participants). fistula and one temporary diplopia) and three participants with
total hippocampectomy (one temporary nerve paresis, two with
Primary outcome: Seizure outcome cerebrospinal fluid subgaleal fistula).
(See also Summary of findings for the main comparison). Primary outcome: Seizure outcome
At one year, 23 out of 40 (58%) of the participants in the surgery (See also Summary of findings 2).
group were free from seizures impairing awareness compared to 3
out of 40 (8%) of the medical treatment group (risk ratio (RR) 7.67, In Liang 2010, at two years, 18 out of 30 (60%) of the participants in
95% confidence interval (CI) 2.50 to 23.51, P value = 0.0004, Analysis the ATL group were free from seizures (Engel Class 1) compared to
1.1). Also at one year, 15 out of 40 (38%) of the participants in the 22 out of 30 (73%) of participants in the aCCT group (RR 1.22, 95%
surgery group were free from all seizures including auras compared CI 0.85 to 1.76, P value = 0.28, Analysis 2.1).
to 1 out of 40 (3%) of the medical treatment group (RR 15.00, 95%
In Schramm 2011, at one year, 77 out of 104 (74%) of the
CI 2.08 to 108.23, P value = 0.007, Analysis 1.2).
participants in the 2.5 cm resection group were free from seizures
The median percentage improvement in monthly seizure frequency (Engel Class 1) compared to 75 out of 103 (73%) of participants in
impairing awareness was 100% in the surgery group compared to the 3.5 cm resection group (RR 1.02, 95% CI 0.86 to 1.20, P value =
34% in the medical treatment group. 0.84, Analysis 2.2).
In Wyler 1995, at one year, 25 out of 36 (69%) of the participants

in the total hippocampectomy group were free from all seizures

Informed decisions.

including auras compared to 13 out of 34 (32%) of the participants 'Extent of resection' (see Analysis 2.3 and 'Extent of resection'
in the partial hippocampectomy group (RR 1.82, 95% CI 1.12 to below). A multivariable regression model of prognostic factors was
2.93, P value = 0.01, Analysis 2.3). Time to first seizure was also also reported in Wyler 1995, see 'Multivariable analyses' below.
reported in Wyler 1995: seizure recurrence occurred earlier in the
partial hippocampectomy group than the total hippocampectomy Overall outcome of surgery according to all randomised and
group (1.3 versus 1.9 years respectively). non-randomised evidence
Out of the 16,508 participants who underwent surgery for epilepsy
Secondary outcome: Seizure outcome according to prognostic
included in the 177 studies, satisfactory seizure outcome data
factors of interest
were available for 16,253 from 173 studies. Among these 16,253
Liang 2010 did not report any univariate or multivariable analyses participants, 10,518 achieved a good outcome (65%), defined as
that investigated the influence of any prognostic factors of interest Engel Class 1, more than one year seizure-free or a good outcome
to us for the seizure outcome. measured using another scale that satisfied our definition of a
good outcome (see Primary outcomes and Included studies). The
Schramm 2011 reported univariate analyses contributing to 'Extent proportion of participants with a good outcome in each study
of resection', 'Presence of mesial temporal sclerosis (MTS)' (by outcome scale) is shown in Figure 2, ranging from 13.5% of
and 'Side of surgical resection' (see below). Results of these participants to 92.5% of participants with a good outcome. We note
univariate analyses were presented for all participants in the that Figure 2 is for illustrative purposes only and the proportion of
study who underwent a surgical procedure and were not reported participants with a good outcome is not pooled in meta-analysis
separated by randomised group. A multivariable regression model in Analysis 3.1 or Figure 2. Further, we note that 'proportion' or
of prognostic factors was also reported in Schramm 2011, see 'difference in proportions' is not an available option for the name of
'Multivariable analyses' below. effect size in Review Manager, therefore in Analysis 3.1 and Figure
2 'standardised mean difference' corresponds to 'proportion with a
The randomised comparison in Wyler 1995 (partial versus total
good outcome of surgery'.
hippocampectomy) contributed to the univariate analysis of

Figure 2. Forest plot of comparison: 1 Surgery for Epilepsy, outcome: 1.1 Proportion with a good outcome of
surgery. We note that 'proportion' or 'difference in proportions' is not an available option for the name of effect

Informed decisions.

size in Review Manager, therefore in Analysis 3.1 and Figure 2 'standardised mean difference (SMD)' corresponds to
'proportion with a good outcome of surgery'.

Informed decisions.


Figure 2. (Continued)

Informed decisions.


Figure 2. (Continued)

See Figure 3 for the funnel plot of the outcome 'Proportion with funnel plot, we do not deem any asymmetry to be present therefore
a good outcome of surgery' (again note that 'SMD' corresponds to we have no evidence that publication bias exists for this outcome.
'proportion') for 174 included studies. From visual inspection of the


Informed decisions.

Figure 3. Funnel plot of comparison: 1 Surgery for Epilepsy, outcome: 1.1 Proportion with a good outcome of
surgery. We note that 'proportion' or 'difference in proportions' is not an available option for the name of effect
size in Review Manager, therefore in Analysis 3.1 and Figure 2 'standardised mean difference (SMD)' corresponds to
'proportion with a good outcome of surgery'.

From further examination of overall post-surgery seizure outcome Recording of adverse events in all included studies
according to site of surgery (166 studies reported these data):
Adverse events were reported in 74 of the 177 (42%) included
722 out of 1253 participants from 14 studies in which participants
studies. Detail was very variable. Some reported deaths only,
underwent extratemporal lobe surgery achieved a good outcome
including those occurring many years later from unrelated causes.
of surgery (62%), 4454 out of 6490 participants from 75 studies in
The timing of events was often not stated, with no clarity around
which participants underwent temporal lobe surgery achieved a
what was peri-operative, what was a transient event (some
good outcome of surgery (69%) and 4942 out of 7554 participants
included persistence of a feature up to 12 months as transient)
from 77 studies in which participants underwent temporal or
and what was a permanent deficit. The overall quality of these
extratemporal lobe surgery achieved a good outcome of surgery
data is therefore very poor. Few studies contained any reference to
(65%). There is a statistically significant association between site of
postoperative cognition or mental state.
surgery and outcome of surgery (Chi2 test P value < 0.001).
Notwithstanding these constraints, adverse events were recorded
For 255 participants from four studies (1.5% of the total 16,575
in 1308 (13.75%) of the 9512 participants involved in these 74
participants undergoing surgery), we did not deem the seizure
studies. Taking the studies where the number with transient
outcome scale to be satisfactory for the outcomes of this review
adverse events were specified (278 (21%)) and adding to this the
and we did not include these participants in any meta-analyses (see
number of events that we can assume to be transient (that is
Primary outcomes and Included studies). Under the definition of
short-lived and treatable, to include infection/fever, cerebrospinal
'good' outcome in these four studies, 142 out of 255 participants
fluid (CSF) leak/collection, haemorrhage, deep venous thrombosis,
achieved a 'good outcome' (56%). These four studies were not
status epilepticus and cerebral oedema) we find that the total
considered further in this review (Kwan 2010; Krsek 2013; Kuzniecky
number with a transient adverse event is 614 (6%). This leaves
1993; Ure 2009).
us 694 of the 9512 (7.3%) with a permanent adverse event. It is
highly likely that this is an over-estimate as a prevalence figure
for permanent neurological deficit as many studies did not record

Informed decisions.

which events were only transient and more than one event could from adjusted multivariable analyses reported as aggregate data,
be recorded in the same person. we were unable to perform any analyses in this review using
adjusted results and we have described the multivariable adjusted
The adverse events recorded are as follows: models including our pre-specified variables of interest below (see
'Multivariate analyses').
Undefined 98 (7.5%); infection/fever (difficult to differentiate
infective causes from autonomic dysfunction) 251 (19.2%); motor We emphasise that the univariate RRs presented in analyses 4.1
impairment (to include mono-, hemi-, facial pareses along with to 4.14 assume that each factor acts independently on the seizure
cranial nerve involvement) 220 (16.8%); visual field defect (13.2%); outcome, which, in reality, is unlikely to be the case. Therefore
haemorrhage 56 (4.3%); language impairment (3.2%); CSF leak we have discussed only the direction of the analyses rather than
or collection (e.g. sub-galeal) 36 (2.8%); cognitive impairment the numerical magnitude and do not recommend the use of these
to include memory loss 34 (2.6); hydrocephalus 24 (1.8%); and results in any future research.
miscellaneous (to include deep venous thrombosis (associated
in three with pulmonary embolism), status epilepticus, cerebral Pre-operative Factors
oedema and urinary incontinence 10 (0.8%). Altered mental state
was recorded in 118 (9%) of participants. The duration is undefined. Results of pre-operative (Magnetic Resonance Imaging) MRI
It is notable that one study contributed 65 of this number as Forty-two studies with 3956 participants reported data on pre-
a result of a detailed psychiatric assessment being part of their operative MRI results (normal versus abnormal, where abnormal is
post-surgery follow-up protocol (Suppiah 2009). A new episode defined within individual studies) as a prognostic factor for seizure
of psychological symptoms occurred in 52% of 114 participants outcome of surgery. Seventeen studies reported seizure outcome
assessed in this way during the first year after surgery. Sixty-six of according to a more than one year seizure-free scale, 21 studies
these people had a lifetime prevalence of anxiety or depression according to the Engel Class Scale and four studies according to
or another axis I disorder. Supplementary publications sometimes other scales (see Included studies for definitions of other scales).
reported on mental state. For example, Cleary 2012 analysed A total of 1252 participants had normal MRI results of which 687
psychiatric diagnosis in 280 of 615 participants reported in de Tisi achieved a good outcome of surgery (55%) and 2074 participants
2011, showing that 38% of 280 had significant psychiatric problems had abnormal MRI results of which 1816 achieved a good outcome
within four years following temporal lobe surgery. of surgery (67%).
The figures given here in brackets are percentages of 1308, the Analysis 4.1 shows that participants with abnormal pre-operative
overall number of recorded adverse events. MRI results are significantly more likely to have a good outcome
of surgery than participants with normal pre-operative MRI results
Prognostic factors of a good outcome of surgery in all included (pooled RR for normal versus abnormal MRI, 95% CI adjusted
studies for outcome scale 0.78 (0.73 to 0.82), P value < 0.00001). Results
Univariate analyses according to outcome scale are very similar: more than one year
seizure-free pooled RR for normal versus abnormal MRI: 0.82 (95%
One hundred and eighteen out of 174 studies reporting seizure CI 0.74 to 0.90), P value < 0.0001, Engel Class Scale pooled RR for
outcome data on a satisfactory scale contributed data towards normal versus abnormal MRI 0.74 (95% CI 0.69 to 0.80), P value <
at least one of our pre-specified prognostic factors of interest 0.00001 and other scale pooled RR for normal versus abnormal MRI
for univariate analysis (i.e. the independent, unadjusted effect 0.84 (95% CI 0.53 to 1.32), P value = 0.45. There is no evidence of
of each of these factors on outcome, see Data extraction and a difference between the outcome scales for this prognostic factor
management). Twenty-nine studies reported one prognostic factor,
(Chi2 test for subgroup differences P value = 0.30).
20 studies reported two prognostic factors, 28 studies reported
three prognostic factors, 16 studies reported four prognostic A moderate amount of heterogeneity is present between studies in
factors, 12 studies reported five prognostic factors, seven studies each of the subgroups and in overall analysis (the I2 value ranges
reported six prognostic factors, two studies reported seven from 20% to 65% in subgroup analyses and is 40% overall). When
prognostic factors and four studies reported nine prognostic the analysis is repeated for subgroups and overall analysis with
factors. See Table 6 for full details of the prognostic factors recorded a random-effects model, the pooled results are similar and the
in each study. conclusions are unchanged. The largest amount of heterogeneity
is present between studies using 'other' scales therefore the
Analyses 4.1 to 4.14 show the univariate risk ratio (RR) and 95%
variability of results here is likely to be due to the different outcome
confidence interval (CI) of good outcome for each prognostic factor.
scales used (see Included studies). Also one study in this subgroup
All analyses are subgrouped according to outcome scales 'more
shows a large effect in favour of an abnormal MRI for a good
than one year seizure-free', 'Engel Class Scale' or 'other scale' and
outcome of surgery (RR 0.17, 95% CI 0.01 to 2.38) as only four
an overall pooled RR and 95% CI adjusted for outcome scale is
participants out of 36 in this study had a normal pre-operative MRI
reported for each factor. All analyses are performed with a fixed-
and none of these participants achieved a good outcome of surgery
effect model unless otherwise stated.
(Sakamoto 2009).
We intended to compare the results of pooling unadjusted
Further subgroup analysis separating site of surgery of participants
univariate analyses to the results of pooling adjusted multivariable
in the study (temporal only, extratemporal only, both temporal
analyses for each prognostic factor, under the assumption that it is
and extratemporal) and study design (prospective or retrospective
likely that our pre-specified factors of interest do not act completely
identification of participants) shows no significant differences
independently on outcome and do in fact interact with each other
between these subgroups, which could be contributing to
as well as the outcome. However, due to limited information

Informed decisions.

heterogeneity (analyses not shown but available from authors). Analysis 4.3 shows that participants with MTS on pathology are
We therefore deduce that the variability between studies may significantly more likely to have a good outcome of surgery than
originate from slightly different definitions of pre-operative MRI participants without MTS on pathology (pooled RR for presence
abnormality across studies; for example, some studies defined versus absence of MTS, 95% CI adjusted for outcome scale: 1.18
only 'abnormalities' on MRI while other studies defined specific (95% CI 1.13 to 1.24), P value < 0.00001). Results according to
abnormalities such as lesions on MRI. Further, small participant outcome scale are: more than one year seizure-free pooled RR for
numbers and unbalanced participant numbers with normal and presence versus absence of MTS: 1.25 (95% CI 1.13 to 1.39), P value
abnormal MRIs leading to large, imprecise results may have <0.0001, Engel Class Scale pooled RR for presence versus absence
contributed to variability between studies (for example, in Adam of MTS: 1.13 (95% CI 1.07 to 1.20), P value < 0.00001 and other scale
1996, out of 30 participants, one had an abnormal MRI and 29 had pooled RR for presence versus absence of MTS: 1.31 (95% CI 1.14 to
a normal MRI). 1.51), P value = 0.0002. There is no statistically significant evidence
of a difference between the outcome scales (Chi2 test for subgroup
Use of pre-operative intracranial (invasive) monitoring
differences P value = 0.06).
Twenty-one studies with 1547 participants reported data on use
of intracranial monitoring (used versus not used) as a prognostic Across subgroups, no heterogeneity was present between studies
factor for seizure outcome of surgery. Six studies reported seizure classifying seizure outcome as more than one year seizure-free (I2 =
outcome according to a more than one year seizure-free scale, 0%). Some heterogeneity was present between studies classifying
14 studies according to the Engel Class Scale and one study seizure outcome by Engel Class Scale (I2 = 30%) and other scales
according to an 'other' scale (see Included studies for definitions (I2 = 17%), so overall heterogeneity between all included studies is
of other scales). A total of 762 participants underwent intracranial quite low (I2 = 28%). As in previous examples, it is likely that the
monitoring of which 448 achieved a good outcome of surgery (59%) small amount of heterogeneity present between studies is due to
and 785 participants did not undergo intracranial monitoring of many studies having small sample sizes and unbalanced numbers
which 564 achieved a good outcome of surgery (72%). of participants with and without MTS. This has led to several large,
imprecise results that may have contributed to variability between
Analysis 4.2 shows that participants who do not undergo studies.
intracranial monitoring are significantly more likely to have a good
outcome of surgery than participants who do undergo intracranial Concordance of pre-operative MRI and EEG
monitoring (pooled RR for intracranial monitoring used versus not
Concordance is where the results of two investigations – usually
used, 95% CI adjusted for outcome scale 0.85 (95% CI 0.78 to 0.93), P
an MRI scan and an EEG – localise the likely source of the epilepsy
value = 0.0002). Results according to outcome scale are: more than
to the same lobe. Twenty-three studies with 1778 participants
one year seizure-free pooled RR for intracranial monitoring used
reported data on concordance of pre-operative MRI and EEG
versus not used: 0.88 (95% CI 0.76 to 1.02), P value = 0.10, Engel
(concordant versus discordant) as a prognostic factor for seizure
Class Scale pooled RR for intracranial monitoring used versus not
outcome of surgery. Eight studies reported seizure outcome
used: 0.85 (95% CI 0.77 to 0.94), P value = 0.004 and other scale
according to a more than one year seizure-free scale, 12 studies
pooled RR for intracranial monitoring used versus not used: 0.53
according to the Engel Class Scale and three studies according
(95% CI 0.28 to 0.98), P value = 0.04. There is no evidence of a
to an 'other' scale (see Included studies for definitions of other
difference between the outcome scales for this prognostic factor
scales). A total of 1200 participants had concordant pre-operative
(Chi2 test for subgroup differences P value = 0.39). MRI and EEG of which 824 achieved a good outcome of surgery
(69%) and 578 participants had discordant pre-operative MRI and
A moderate amount of heterogeneity is present between studies
EEG of which 313 (54%) achieved a good outcome of surgery.
in each of the subgroups and in overall analysis (the I2 value
ranges from 37% to 46% in subgroup analyses and is 37% overall). Analysis 4.4 shows that participants with concordant pre-operative
When the analysis is repeated with a random-effects model the MRI and EEG are significantly more likely to have a good outcome
pooled results are similar and the conclusions are unchanged. We of surgery than participants with discordant pre-operative MRI and
assume that it is likely that the small amount of heterogeneity EEG (pooled RR for concordant versus discordant, 95% CI adjusted
present between studies is due to small participant numbers and for outcome scale 1.25 (95% CI 1.15 to 1.37), P value < 0.00001).
unbalanced participant numbers having or not having intracranial Results according to outcome scale are: more than one year seizure-
monitoring used leading to large, imprecise results that may have free pooled RR for concordant versus discordant: 1.21 (95% CI 1.07
contributed to variability between studies. to 1.37), P value = 0.003, Engel Class Scale pooled RR for concordant
versus discordant: 1.27 (95% CI 1.11 to 1.46), P value = 0.0005 and
Mesial temporal sclerosis (MTS) on MRI or pathology
other scale pooled RR for concordant versus discordant: 1.40 (95%
Forty-six studies with 4354 participants reported data on mesial CI 1.02 to 1.93), P value = 0.04. There is no evidence of a difference
temporal sclerosis on pathology (present versus absent) as a between the outcome scales for this prognostic factor (Chi2 test for
prognostic factor for seizure outcome of surgery. Nine studies subgroup differences P value = 0.65).
reported seizure outcome according to a more than one year
seizure-free scale, 30 studies according to the Engel Class Scale No heterogeneity is present between studies classifying seizure
and seven studies according to an 'other' scale (see Included outcome according to the Engel Class Scale (I2 = 0%). A moderate
studies for definitions of other scales). A total of 1735 participants amount of heterogeneity is present between studies in other
had confirmed MTS on pathology of which 1287 achieved a good subgroups (the I2 value is 50% to 56% and I2 = 26% overall). When
outcome of surgery (74%) and 2619 participants did not show MTS the analysis is repeated with a random-effects model, the pooled
on pathology of which 1609 (62%) achieved a good outcome of RR for concordant versus discordant is no longer significant in the
surgery.
Informed decisions.

subgroups more than one year seizure-free scale (1.24, 95% CI 0.97 to an 'other' scale (see Included studies for definitions of other
to 1.59, P value = 0.08) and other scale (1.46, 95% CI 0.89 to 2.39, scales). A total of 440 participants had a history of febrile seizures
P value = 0.14). However, the overall pooled result adjusted for of which 343 achieved a good outcome of surgery (78%) and 928
outcome scale is similar and the conclusions remain unchanged participants had no history of febrile seizures of which 615 (66%)
(pooled RR for concordant versus discordant 1.20, 95% CI 1.08 to achieved a good outcome of surgery.
1.34, P value = 0.0009).
Analysis 4.5 shows that participants with a history of febrile seizures
Heterogeneity present between studies using 'other' scales is likely are significantly more likely to have a good outcome of surgery
to be due to the three different outcome scales used by the three than participants without a history of febrile seizures (pooled RR
studies (Holmes 2000; Rossi 1994; Yang 2011, see Included studies). for history versus no history of febrile seizures, 95% CI adjusted for
From inspection of the studies that classified seizure outcome as outcome scale: 1.09 (95% CI 1.01 to 1.17), P value = 0.002). Results
more than one year seizure-free, small participant numbers and according to outcome scale are: more than one year seizure-free
unbalanced participant numbers with concordant and discordant pooled RR for history versus no history of febrile seizures: 1.18 (95%
pre-operative MRI and EEG leading to large, imprecise results may CI 1.05 to 1.32), P value = 0.006, Engel Class Scale pooled RR for
have contributed to variability between studies. Further, three history versus no history of febrile seizures: 1.01 (95% CI 0.92 to
studies show particularly variable results. Tatum 2008 shows a large 1.11), P value = 0.83 and other scale (one study) RR for history versus
significant effect in favour of concordance, with wide confidence no history of febrile seizures: 1.11 (95% CI 0.90 to 1.37), P value
intervals; participants in this study had a normal MRI for inclusion = 0.32. There is no evidence of a difference between the outcome
therefore concordance with MRI was determined operatively rather scales for this prognostic factor (Chi2 test for subgroup differences
than pre-operatively. Kim 2010b also showed a large effect in favour P value = 0.14).
of concordance, with wide confidence intervals; all participants
in this study had dual pathology of MTS and FCD (focal cortical A substantial amount of heterogeneity is present between studies
dysplasia) for inclusion. Kim 2009 is the only study to show an effect that classified seizure outcome according to more than one year
in favour of discordance; all participants in this study had FCD for seizure-free (I2 = 70%) and no heterogeneity is present between the
inclusion. Other studies in this subgroup all showing moderate, other outcome scales (I2 = 0%), which leads to an overall moderate
non-significant effects in favour of concordance had no specific amount of heterogeneity in the overall analysis (I2 = 32%). When
inclusion criteria based on imaging or pathology. It is feasible that the analysis is repeated with a random-effects model, the pooled
specific study inclusion criteria based on how the pathological RR is unchanged in the subgroups without heterogeneity between
lesions are sited may in turn influence the classification of the pre- studies but in the subgroup more than one year seizure-free, and
operative MRI and therefore the concordance with pre-operative now in the overall analysis, the advantage for participants with a
EEG (see below for the apparent disadvantage of retrospective history of febrile seizures is no longer statistically significant (more
design in this respect). than one year seizure-free pooled RR for history versus no history of
febrile seizures: 1.23 (95% CI 0.96 to 1.57), P value = 0.10) and overall
Further subgroup analysis separating the site of surgery pooled RR for history versus no history of febrile seizures: 1.07 (95%
of participants in the study (temporal only, extratemporal CI 0.98 to 1.17), P value = 0.12).
only, both temporal and extratemporal) shows no significant
difference between these subgroups that could be contributing to From inspection of the studies that classified seizure outcome
heterogeneity (analyses not shown but available from authors). as more than one year seizure-free small participant numbers
Subgroup analysis according to study design (prospective or and unbalanced participant numbers with and without a history
retrospective identification of participants) shows a larger of febrile seizures leading to large, imprecise results may have
advantage for concordant pre-operative MRI and EEG in the contributed to the variability between studies as in previous
three studies of a prospective design than the 20 studies of a analyses, particularly the two smallest studies showing the two
retrospective design (prospective pooled RR for concordant versus largest effects in favour of a history of febrile seizures (Holmes 1997;
discordant: 1.91 (95% CI 1.06 to 3.44), P value = 0.03, retrospective Kim 2010b).
pooled RR for concordant versus discordant: 1.21 (95% CI 1.10 to
1.33), P value < 0.0001). There is no statistically significant evidence A potential association between febrile seizures and mesial
of difference between subgroups (Chi2 test for subgroup differences temporal sclerosis, particularly whether prolonged febrile seizures
P value = 0.13); however, this trend may be contributing towards the are a risk factor for febrile seizures, is well documented (Davis 1996;
heterogeneity. For example, in studies with a prospective design Maher 1995; Sarkisian 1999; Scott 2002; Szabo 1999). Therefore as
determining concordance of tests performed during the study may we have already noted in this review a good outcome associated
be easier and more reliable than determining concordance from with the presence of MTS on pathological examination (see Analysis
medical records of tests that have been previously performed in a 4.3), it is intuitive that we should also observe a good outcome
study of a retrospective design. In summary, it is likely that study of surgery for participants with a history of febrile seizures. Seven
characteristics such as design, inclusion criteria and outcome scale studies have recorded data on the presence of both MTS on
may have contributed to the variability for this prognostic factor. pathological examination and history of febrile seizures: two show
a significant advantage for participants with MTS (Radhakrishnan
History of febrile seizures 1998; Spencer 2005), three show a non-significant advantage
Fifteen studies with 1368 participants reported data on history of for MTS (Jeong 1999; Perry 2010; Terra-Bustamante 2005a), and
febrile seizures (history versus no history) as a prognostic factor two show a small non-significant advantages for no MTS on
for post-surgery seizure outcome. Five studies reported seizure pathological examination (Chabardes 2005; Grivas 2006). The same
outcome according to a more than one year seizure-free scale, nine trends are seen in the analysis of febrile seizures as a prognostic
studies according to the Engel Class Scale and one study according factor; the five studies that show an advantage for MTS on

Informed decisions.

pathological examination (significant or non-significant) also show Scale and one study according to an 'other' scale (see Included
an advantage for a history of febrile seizures in relation to a studies for definitions of other scales). A total of 35 participants
good post-surgery outcome; and the two studies that show small had encephalomalacia of which nine achieved a good outcome of
advantages for no MTS also show small advantages for no history of surgery (25%) and 239 participants did not have encephalomalacia
febrile seizures (see Analysis 4.5). of whom 90 achieved a good outcome of surgery (38%).
Given this apparent association between febrile seizures and MTS Analysis 4.7 shows that there is no significant difference
pathology, as in Analysis 4.4, specific inclusion criteria of studies between the presence or absence of encephalomalacia on the
based on pathology may be contributing to the variability observed outcome of surgery (pooled RR for presence versus absence of
in the subgroup of studies that classified seizure outcome by a more encephalomalacia, 95% CI adjusted for outcome scale 0.67 (95%
than one year seizure-free scale in Analysis 4.5. Two studies in this CI 0.37 to 1.21), P value = 0.18). In other words, presence of
subgroup included only participants with a diagnosis of MTS (Walz encephalomalacia on pathology is not a significant independent
2003), or a diagnosis of dual pathology MTS and FCD (Kim 2010b), predictor of seizure outcome of surgery. Results according to
and one study excluded participants with any pathology other than outcome scale are: Engel Class Scale pooled RR for presence versus
FCD including MTS (Kim 2009). The two remaining studies had no absence of encephalomalacia: 0.82 (95% CI 0.45 to 1.50), P value =
specific inclusion criteria based on pathological findings (Holmes 0.52 and other scale (one study) RR for presence versus absence of
1997; Spencer 2005). encephalomalacia: 0.28 (95% CI 0.04 to 1.87), P value = 0.19. There
is no evidence of a difference between the outcome scales for this
History of head injury
prognostic factor (Chi2 test for subgroup differences P value = 0.29).
Seven studies with 551 participants reported data on history of There is no evidence of heterogeneity between studies in any of the
head injury (history versus no history) as a prognostic factor subgroup analyses or overall (I2 = 0%).
for post-surgery seizure outcome. Two studies reported seizure
outcome according to a more than one year seizure-free scale, Focal cortical dysplasia/malformation of cortical development on
pathology
three studies according to the Engel Class Scale and two studies
according to an 'other' scale (see Included studies for definitions Forty-five studies with 3529 participants reported data on a focal
of other scales). A total of 159 participants had a history of head cortical dysplasia (FCD)/malformation of cortical development
injury of which 100 achieved a good outcome of surgery (63%) and (MCD) on pathological examination (present versus absent, where
392 participants had no history of head injury of which 242 (62%) defect was generally defined as cortical dysplasia or malformation
achieved a good outcome of surgery. of cortical development) as a prognostic factor for seizure outcome
of surgery. Eight studies reported seizure outcome according to
Analysis 4.6 shows that there is no significant difference between a more than one year seizure-free scale, 32 studies according to
history or no history of head injury on the outcome of surgery the Engel Class Scale and five studies according to an 'other' scale
(pooled RR for history versus no history of head injury, 95% CI (see Included studies for definitions of other scales). A total of
adjusted for outcome scale 0.99 (95% CI 0.86 to 1.13), P value = 1190 participants showed a focal cortical dysplasia/malformation
0.85). Results according to outcome scale are: more than one year of cortical development of which 672 achieved a good outcome of
seizure-free pooled RR for history versus no history of head injury: surgery (56%) and 2339 participants did not show a focal cortical
0.87 (95% CI 0.72 to 1.05), P value = 0.14, Engel Class Scale pooled dysplasia/malformation of cortical development of which 1584
RR for history versus no history of head injury: 1.17 (95% CI 0.99 to achieved a good outcome of surgery (68%).
1.37), P value = 0.06 and other scale pooled RR for history versus no
history of head injury: 0.83 (95% CI 0.51 to 1.33), P value = 0.43. Analysis 4.8 shows that participants without a focal cortical
dysplasia/malformation of cortical development are significantly
Here, there is evidence of a difference between the outcome scales more likely to have a good outcome of surgery than
(Chi2 test for subgroup differences P value = 0.05), where the trend participants with a focal cortical dysplasia/malformation of cortical
is that studies classifying outcome on the Engel Class Scale favour development on pathology (pooled RR for presence versus absence
a history of head injury for a good outcome of surgery, while of FCD/MCD, 95% CI adjusted for outcome scale: 0.90 (95% CI 0.85
studies classifying outcome on a more than one year seizure-free to 0.95), P value = 0.0005). Results according to outcome scale are:
scale or other scale favour no history of head injury. Heterogeneity more than one year seizure-free pooled RR for presence versus
is present only between studies using other scales to classify absence of FCD/MCD: 0.92 (95% CI 0.84 to 1.02), P value = 0.11, Engel
seizure outcome (I2 = 71%), which is the source of the moderate Class Scale pooled RR for presence versus absence of FCD/MCD:
heterogeneity in the analysis overall (I2 = 46%). When the analysis 0.88 (95% CI 0.82 to 0.96), P value = 0.003 and other scale pooled
is repeated for subgroups and the overall analysis with a random- RR for presence versus absence of FCD/MCD: 0.91 (95% CI 0.73 to
effects model, the pooled results are similar and the conclusions 1.13), P value = 0.39. There is no evidence of a difference between
are unchanged. It is likely that this variation is due to the two the outcome scales for this prognostic factor (Chi2 test for subgroup
different 'other' scales used by the studies. Holmes 2000 classified differences P value = 0.82).
more than two years seizure-free as a good outcome, while Yang
2011 classed ILAE classes 1 and 2 as a good outcome. A moderate amount of heterogeneity is present between studies
in each of the subgroups and in the overall analysis: the I2 value
Encephalomalacia on pathology ranges from 33% to 41% in subgroup analyses and is 31% overall.
Four studies with 274 participants reported data on When the analysis is repeated with a random-effects model the
encephalomalacia on pathological examination (present versus pooled results are similar and the conclusions are unchanged. As
absent) as a prognostic factor for seizure outcome of surgery. Three in previous analyses, we assume that it is likely that the small
studies reported seizure outcome according to the Engel Class amount of heterogeneity present between studies is due to small

Informed decisions.

participant numbers and unbalanced participant numbers with Vascular malformation on pathology
and without a focal cortical dysplasia/malformation of cortical Nineteen studies with 1488 participants reported data on
development leading to large, imprecise results that may have vascular malformations (present versus absent, where specific
contributed to variability between studies. Furthermore, variability malformations were defined in some of the included studies) as
across studies in the exact definition of the focal cortical dysplasia/ a prognostic factor for post-surgery outcome. One study reported
malformation of cortical development (focal cortical dysplasia or seizure outcome according to a more than one year seizure-free
malformation of cortical development of varying severities) may scale, 13 studies according to the Engel Class Scale and five studies
have also contributed to variability in results between studies. according to an 'other' scale (see Included studies for definitions of
Tumour on pathology other scales). A total of 139 participants had a confirmed vascular
malformation of whom 89 achieved a good outcome post-surgery
Forty-one studies with 3357 participants reported data on tumour (64%) and 1349 participants did not show a vascular malformation
as a pathological finding (present versus absent, where specific of whom 785 achieved a good post-surgery outcome (58%).
tumour types were defined in some of the included studies) as
a prognostic factor for seizure outcome of surgery. Seven studies Analysis 4.10 shows that there is no significant difference between
reported seizure outcome according to a more than one year presence or absence of vascular malformation on the outcome
seizure-free scale, 28 studies according to the Engel Class Scale of surgery (pooled RR for presence versus absence of vascular
and six studies according to an 'other' scale (see Included studies malformation, 95% CI adjusted for outcome scale: 1.07 (95% CI 0.94
for definitions of other scales). A total of 806 participants had to 1.21), P value = 0.34). In other words, presence of a vascular
a confirmed tumour of which 595 achieved a good outcome of malformation on pathology is not a significant independent
surgery (74%) and 2551 participants did not show a tumour of predictor of seizure outcome of surgery. Results according to
which 1512 achieved a good outcome of surgery (59%). outcome scale are very similar: more than one year seizure-free
pooled RR for presence versus absence of vascular malformation:
Analysis 4.9 shows that participants with a tumour are significantly 1.06 (95% CI 0.62 to 1.79), P value = 0.84, Engel Class Scale pooled RR
more likely to have a good outcome of surgery than participants for presence versus absence of vascular malformation: 1.14 (95% CI
without a tumour (pooled RR for presence versus absence of 0.98 to 1.34), P value = 0.09 and other scale pooled RR for presence
tumour, 95% CI adjusted for outcome scale: 1.21 (95% CI 1.15 versus absence of vascular malformation: 1.07 (95% CI 0.94 to 1.21),
to 1.28), P value < 0.00001). Results according to outcome scale P value = 0.64. There is no evidence of a difference between the
are very similar: more than one year seizure-free pooled RR for outcome scales for this prognostic factor (Chi2 test for subgroup
presence versus absence of tumour: 1.16 (95% CI 1.03 to 1.31), differences P value = 0.42). There is no evidence of heterogeneity
P value = 0.02, Engel Class Scale pooled RR for presence versus between studies in any of the subgroup analyses or overall (I2 = 0%).
absence of tumour: 1.20 (95% CI 1.13 to 1.29), P value < 0.00001
and other scale pooled RR for presence versus absence of tumour: Distribution of interictal spikes
1.34 (95% CI 1.12 to 1.60), P value = 0.001. There is no evidence of
a difference between the outcome scales for this prognostic factor Eighteen studies with 1404 participants reported data on
distribution of interictal spikes (unilateral versus bilateral spikes,
(Chi2 test for subgroup differences P value = 0.41).
also defined as lateralising versus non-lateralising spikes, focal
A moderate amount of heterogeneity is present between studies versus non-focal spikes or localising versus non-localising spikes)
in each of the subgroups and in the overall analysis: the I2 value as a prognostic factor for seizure outcome of surgery. Seven studies
ranges from 34% to 54% and is 41% overall. The largest amount reported seizure outcome according to a more than one year
of heterogeneity is present between studies classifying seizure seizure-free scale, six studies according to the Engel Class Scale and
five studies according to an 'other' scale (see Included studies for
outcome by more than one year seizure-free (I2 = 54%), however the
definitions of other scales). A total of 722 participants had unilateral
variability is greatly influenced by a single study in this subgroup
spikes of which 504 achieved a good outcome of surgery (70%) and
with a large effect size due to all participants in this study with
682 participants has bilateral spikes of which 406 achieved a good
a tumour achieving a good outcome of surgery (Duchowny 1998).
outcome of surgery (59%).
When the analysis is repeated with a random-effects model in
this subgroup, the pooled RR for presence versus absence of Analysis 4.11 shows that participants with unilateral interictal
tumour is no longer statistically significant: 1.17 (95% CI 0.97 to spikes are significantly more likely to achieve a good outcome of
1.41), P value = 0.10. Within other subgroups and in the overall surgery than participants with bilateral interictal spikes (pooled RR
analysis, the pooled results are similar and the conclusions are for unilateral versus bilateral spikes, 95% CI adjusted for outcome
unchanged. As in the previous analyses, we assume that it is scale: 1.14 (95% CI 1.05 to 1.24), P value < 0.0001). Results according
likely that the small amount of heterogeneity present between to outcome scale are: more than one year seizure-free pooled RR for
studies is due to small participant numbers and unbalanced unilateral versus bilateral spikes: 1.08 (95% CI 0.94 to 1.24), P value
participant numbers with and without tumours leading to large, = 0.72, Engel Class Scale pooled RR for unilateral versus bilateral
imprecise results contributing to variability between studies. spikes: 1.19 (95% CI 1.04 to 1.36), P value = 0.009 and other scale
Furthermore, variability in the type of tumour observed across pooled RR for unilateral versus bilateral spikes: 1.16 (95% CI 0.97 to
studies (i.e. some studies recorded the specific types of tumours 1.39), P value = 0.11. There is no evidence of a difference between
that were observed, such as dysembryoplastic neuroepithelial the outcome scales for this prognostic factor (Chi2 test for subgroup
tumours (DNET), ganglioglioma, oligodemdroglioma etc., while differences P value = 0.58).
other studies reported that non-specific 'tumours' were observed)
may have also contributed to variability in results between studies. A substantial amount of heterogeneity is present in the analysis
overall (I2 = 67%) and between studies that classified seizure

Informed decisions.

outcome according to the Engel Class Scale (I2 = 84%) and other of other scales). A total of 1716 participants underwent complete
scales (I2 = 82%). No heterogeneity is present between studies surgical resection of which 1277 achieved a good outcome of
that classified seizure outcome according to a more than one year surgery (74%) and 1297 participants underwent a less complete
seizure-free scale (I2 = 0%). When the analyses are repeated with surgical resection of which 725 achieved a good outcome of surgery
a random-effects model for studies that classified seizure outcome (56%).
according to the Engel Class Scale the pooled RR for unilateral
Analysis 4.12 shows that participants with complete surgical
versus bilateral spikes is no longer statistically significant: 1.33
resection are significantly more likely to achieve a good outcome
(95% CI 0.88 to 2.00), P value = 0.17. Results for other subgroups
of surgery than participants with a less complete resection (pooled
and the overall analysis are similar and the conclusions remain
RR for complete versus incomplete resection, 95% CI adjusted for
unchanged.
outcome scale 1.41 (95% CI 1.32 to 1.50), P value < 0.00001). Results
As in the previous analyses, the large amount of heterogeneity according to outcome scale are: more than one year seizure-free
is present between studies using 'other' scales. Therefore the pooled RR for complete versus incomplete resection: 2.00 (95% CI
variability of results may be due to the different outcome scales 1.66 to 2.41), P value < 0.00001, Engel Class Scale pooled RR for
used (see Included studies). Small participant numbers and complete versus incomplete resection: 1.29 (95% CI 1.21 to 1.39),
unbalanced participant numbers with unilateral and bilateral P value < 0.00001 and other scale pooled RR for complete versus
spikes leading to large, imprecise results may have contributed to incomplete resection: 1.59 (95% CI 1.15 to 2.20), P value = 0.005.
variability between studies. There is highly significant evidence of a difference between the
outcome scales for this prognostic factor (Chi2 test for subgroup
Further subgroup analysis separating the site of surgery of differences P value < 0.0001); a larger advantage is present for
participants in the study (temporal only, extratemporal only, both studies that classify studies by a more than one year seizure-free
temporal and extratemporal) and study design (all studies are of scale than the other two outcome scales.
a retrospective design) shows no significant differences between
these subgroups that could be contributing to heterogeneity A substantial amount of heterogeneity is present between all
(analyses not shown but available from authors). subgroups and overall in the analysis (the I2 value ranges from
67% to 86% in subgroups and is 77% in the overall analysis). When
We note that data on interictal spikes are defined in slightly the analyses are repeated with a random-effects model for studies
different ways across studies: two studies defined lateralising which classify seizure outcome according to 'other scales' the
versus non-lateralising or contralateral spikes (Boshuisen 2010; pooled RR for complete versus incomplete resection is no longer
Greiner 2011), three studies defined localised versus non-localised statistically significant: 1.15, 95% CI 0.44 to 3.00, P value = 0.78.
spikes (Dalmagro 2005; Lee 2008; Tatum 2008), four studies defined Results for the other outcome scale and in the overall analysis are
focal versus non-focal spikes (Jayakar 2008; Kim 2010b; Kim 2009; similar and the results are unchanged.
Rossi 1994), and nine studies defined unilateral versus bilateral
spikes (Chee 1993; Erickson 2005; Goldstein 1996; Holmes 2000; As in previous analyses, some of the variability present between
Madhavan 2007; Remi 2011; Walz 2003; Weinand 1992). Using these studies using Engel Class Scale may be due to small participant
differing definitions as subgroups: lateralising spikes pooled RR numbers and unbalanced participant numbers with unilateral and
for lateralising versus non-lateralising spikes: 1.05 (95% CI 0.85 bilateral spikes leading to large, imprecise results. Also, some of
to 1.31), P value = 0.65, I2 = 11%; localising spikes pooled RR for the heterogeneity between studies using 'other' scales may be
localised versus non-localised spikes: 1.08 (95% CI 0.81 to 1.44), due to the different outcome scales used (see Included studies);
P value = 0.59, I2 = 0%; focal spikes pooled RR for focal versus two studies in this subgroup classified according to the ILAE scale
non-focal spikes: 1.05 (95% CI 0.85 to 1.30), P value = 0.65, I2 (Lackmayer 2013; Sakamoto 2009), and show a non-significant
= 11%; unilateral spikes pooled RR for unilateral versus bilateral trend in favour of less complete resection, while the other study
that shows a large effect in favour of complete resection classified
spikes: 1.22 (95% CI 1.10 to 1.34), P value < 0.00001, I2 = 76%.
according to more than three years seizure freedom (Rossi 1994).
There is no statistically significant evidence of a difference between
these subgroups (Chi2 test for subgroup differences P value = 0.39), We carried out further subgroup analysis separating the site of
however a trend is apparent for a larger effect in favour of unilateral surgery of participants in the study (temporal only (13 studies),
spikes when compared with bilateral spikes as opposed to the other extratemporal only (one study), both temporal and extratemporal
definitions of data relating to the distribution of interictal spikes. (24 studies), site of surgery not available for Dalmagro 2005 and
We deduce that the definition of the prognostic factor is likely to Raabe 2012). This shows the following results (see Analysis 4.13):
have influenced this analysis and therefore we recommend caution extratemporal only RR 2.00 (95% CI 0.76 to 5.29), P value = 0.16;
when interpreting the results of this analysis. temporal only pooled RR for complete versus incomplete resection
1.11 (95% CI 1.03 to 1.20), P value = 0.006, I2 = 31%; temporal and
Operative factors
extratemporal pooled RR for complete versus incomplete resection
Extent of surgical resection 1.98 (95% CI 1.77 to 2.23), P value < 0.00001, I2 = 75%; and overall
pooled RR 1.48 (95% CI 1.38 to 1.58), P value < 0.00001, I2 = 78%.
Forty studies with 3013 participants reported data on 'Extent
There is highly significant evidence of a difference between the
of surgical resection' (complete versus less complete, where
completeness is defined in individual studies) as a prognostic site of surgery and the outcome of surgery (Chi2 test for subgroup
factor for seizure outcome of surgery. Nine studies reported seizure differences P value < 0.00001); there is less of an advantage for
outcome according to a more than one year seizure-free scale, complete surgical resection in studies in which participants had
28 studies according to the Engel Class Scale and three studies temporal lobe surgery only compared to the study in which all
according to an 'other' scale (see Included studies for definitions
Informed decisions.

participants had extratemporal lobe surgery or the studies in which or other scales (I2 = 0%), however some heterogeneity is present
participants had either temporal or extratemporal lobe surgery. between studies classifying seizure outcome according to the Engel
Class Scale (I2 = 43%) and in the overall analysis (I2 = 28%). When
Furthermore, the majority of the heterogeneity in this analysis is the analyses are repeated with a random-effects model for studies
between the studies in which participants had either temporal or that classified seizure outcome according to the Engel Class Scale
extratemporal lobe surgery. This could be due to the association the pooled RR and overall pooled RR are no longer statistically
we observed between site of surgery and seizure outcome of significant: Engel Class Scale pooled RR for left versus right-sided
surgery (see 'Overall outcome of surgery' above). In other words, resection: 0.96 (0.89 to 1.02), P value = 0.19, overall pooled RR for
the difference in site of surgery across studies is confounding the left versus right-sided resection: 0.96 (95% CI 0.91 to 1.01), P value
analysis of extent of resection. Analysis 4.13 could also suggest = 0.14. There is no evidence of a difference between the outcome
that there may be a difference in the feasibility of performing a
scales for this prognostic factor (Chi2 test for subgroup differences
complete or less complete resection for temporal or extratemporal
P value = 0.60).
lobes; therefore it would be expected that results would be variable
between studies that include a mixture of temporal lobe and As in previous analyses, the heterogeneity that is present between
extratemporal lobe surgery candidates and outcomes may be studies using the Engel Class Scale may be due to small participant
dependent on the proportion of participants receiving each type of numbers and unbalanced participant numbers with left and right-
surgery. sided resections leading to large, imprecise results.
In addition, we note that the definition of a 'complete' or We carried out further subgroup analysis separating the site of
'less complete' resection was variable across studies. The surgery of participants in the study (temporal only (30 studies),
majority of studies defined the extent of resection by the extratemporal only (two studies), both temporal and extratemporal
type of surgery performed; e.g. anterior temporal lobectomy (four studies), site of surgery not available for Dalmagro 2005).
or extended resection is complete resection while selective This showed the following results (see Analysis 4.15): extratemporal
amygdalohippocampectomy or lesionectomy is less complete only pooled RR for left versus right-sided resection: 1.03 (95% CI
resection (for example, Lackmayer 2013; Sakamoto 2009). Other 0.76 to 1.39), P value = 0.84, I2 = 85%; temporal only pooled RR
studies defined complete or incomplete resection postoperatively for left versus right-sided resection: 0.93 (95% CI 0.89 to 0.98), P
by MRI (for example, O'Brien 2000; Zentner 1996), or operatively
value = 0.006, I2 = 30%; temporal and extratemporal pooled RR
by subdural electroencephalography (EEG) (for example, Widdess-
for left versus right-sided resection: 0.98 (95% CI 0.84 to 1.14),
Walsh 2007). In other studies, the completeness of tailored
P value = 0.77, I2 = 0%; and overall pooled RR for left versus
resection was determined by the surgical team at the time (for
right-sided resection: 0.94 (95% CI 0.90 to 0.98), P value = 0.009,
example, Hamiwka 2005). These differences in definition are also
likely to have contributed to the substantial variability in this I2 = 28%. There is no statistically significant difference between
analysis. sites of surgery for this prognostic factor (Chi2 test for subgroup
differences P value = 0.68), however this subgroup analysis shows
Side of surgical resection that the significant advantage of right-sided resection for good
outcome of surgery exists only in studies in which participants all
Thirty-seven studies with 2976 participants reported data on side of
had temporal lobe surgery. Further, this subgroup analysis shows
surgical resection (left versus right-sided resection) as a prognostic
that the two studies with the largest effects favouring left and right
factor for post-surgery seizure outcome. Five studies reported
resection respectively (Lee 2008; Liava 2012), which contribute to
seizure outcome according to a more than one year seizure-free
the majority of variability in this analysis, are the only two studies
scale, 27 studies according to the Engel Class Scale and five studies
in which all participants had extratemporal lobe surgery.
according to an 'other' scale (see Included studies for definitions
of other scales). A total of 1749 participants underwent left- In summary, the results of this analysis suggest that there is likely to
sided surgical resection of which 1302 achieved a good outcome be an association between side of surgery and outcome of surgery
of surgery (67%) and 1476 participants underwent a right-sided (right-sided resection is associated with good outcome), however
surgical resection of which 1056 achieved a good outcome of the results may have been confounded by the site of surgery
surgery (72%). and this association may only exist for participants undergoing
temporal lobe surgery. We require more evidence on the side
Analysis 4.14 shows that participants with right-sided resection
of extratemporal surgery related to seizure outcome to make
are significantly more likely to achieve a good outcome of surgery
conclusions regarding side of resection as a prognostic factor for
than participants with left-sided resection (pooled RR for left versus
outcome.
right-sided resection, 95% CI adjusted for outcome scale 0.94 (95%
CI 0.90 to 0.98), P value = 0.008). Results according to outcome scale Post-operative factors
are: more than one year seizure-free pooled RR for left versus right-
sided resection: 1.01 (95% CI 0.90 to 1.13), P value = 0.90, Engel Class Postoperative discharges
Scale pooled RR for left versus right-sided resection: 0.93 (95% CI Six studies with 542 participants reported data on postoperative
0.88 to 0.98), P value = 0.01 and other scale pooled RR for left versus discharges (present versus absent) as a prognostic factor for post-
right-sided resection: 0.92 (95% CI 0.79 to 1.07), P value = 0.27. There surgery seizure outcome. Two studies reported seizure outcome
is no evidence of a difference between the outcome scales for this according to a more than one year seizure-free scale, four studies
prognostic factor (Chi2 test for subgroup differences P value = 0.47). according to the Engel Class Scale and one study according to an
'other' scale (see Included studies for definitions of other scales). A
No heterogeneity is present between studies classifying seizure total of 200 participants had postoperative discharges of which 132
outcome according to a more than one year seizure-free scale achieved a good outcome of surgery (66%) and 342 participants did

Informed decisions.

not have postoperative discharges of which 262 achieved a good The results of these further analyses and our observation of
outcome of surgery (77%). an association between site of surgery and outcome of surgery
(see 'Overall outcome of surgery' above) show that absence of
Analysis 4.16 shows that there is no significant difference between postoperative discharges may be a predictor of good outcome of
presence or absence of postoperative discharges in the outcome of surgery. However, this analysis is likely to have been confounded
surgery (pooled RR for presence versus absence of postoperative by differing sites of surgery across the included studies and we
discharges, 95% CI adjusted for outcome scale: 0.91 (95% CI 0.79 to require more evidence on this prognostic factor specific to the
1.04), P value = 0.16). In other words, the presence of postoperative site of surgery to make any conclusions. We note the limitation
discharges is not a significant independent predictor of seizure in this analysis that the factor 'postoperative discharges' may
outcome of surgery. Results according to outcome scale are: more be related to and influenced by other pre and postoperative
than one year seizure-free pooled RR for presence versus absence factors and potentially to outcome; for example, whether or
of postoperative discharges: 1.19 (95% CI 0.82 to 1.73), P value = not a postoperative EEG is obtained, which may be more
0.36, Engel Class Scale pooled RR for presence versus absence of likely if an individual has experienced a possible postoperative
postoperative discharges: 0.85 (95% CI 0.74 to 0.98), P value = 0.03. seizure. Furthermore, postoperative EEGs may be more likely
There is no evidence of a difference between the outcome scales for to be performed after a long interval following surgery to
this prognostic factor (Chi2 test for subgroup differences P value = inform decisions such as antiepileptic drug tapering. Selective
0.10). use of postoperative EEGs and therefore selective participants
contributing to this factor may have introduced selection bias
A substantial amount of heterogeneity is present in the analysis into this analysis, therefore we recommend extreme caution when
overall (I2 = 73%) and between the two studies that classified interpreting the results of this analysis.
seizure outcome according to a more than one year seizure-free
scale (Jennum 1993; Widdess-Walsh 2007) (I2 = 90%), the two Multivariable analyses
studies that show the largest effects in favour of discharges present Summary of studies reporting multivariable models
and absent, respectively. Further, some heterogeneity is present
between studies that classified seizure outcome according to the We screened all 174 studies reporting seizure outcome data
Engel Class Scale (I2 = 34%). When the analyses are repeated with on a satisfactory scale in an additional post hoc analysis. This
a random-effects model, the results are as follows: overall pooled determined whether a multivariable prognostic regression model
RR adjusted for outcome scale: 0.91 (95% CI 0.68 to 1.22), P value = had been fitted to assess the adjusted influence of independent
0.52; more than one year seizure-free pooled RR for presence versus prognostic factors of interest in this review (see Secondary
absence of postoperative discharges: 0.97 (95% CI 0.24 to 3.93), outcomes) on the dependent variable of seizure outcome (defined
P value = 0.97, Engel Class Scale pooled RR for presence versus as the proportion of individuals experiencing a good outcome of
absence of postoperative discharges: 0.85 (95% CI 0.70 to 1.03), P surgery, see Primary outcomes for further details on this definition).
value = 0.09.
In 29 out of 174 studies (17%), no statistical analysis was performed
Further subgroup analysis separating the site of surgery of at all and only proportions/percentages were reported. In 72 out of
participants in the study (temporal only (Janszky 2003b; 174 studies (41%) only univariate analyses were described. In 19 out
Miserocchi 2013; Radhakrishnan 1998) versus both temporal and of 174 studies (11%) seizure outcome was not reported according
extratemporal (Jennum 1993; Widdess-Walsh 2007; Wray 2012)) to our definitions of a 'good' seizure outcome for analyses involving
shows a significant difference between these subgroups (Chi2 test prognostic factors. In 16 out of 174 studies (9%), a multivariable
for subgroup differences P value = 0.02, see Analysis 4.17). Within model was described but did not include any of our pre-specified
the three studies in which temporal lobe surgery is performed factors.
on participants, the pooled RR for presence versus absence of
In total, 29 studies reported a multivariable prognostic model
postoperative discharges is 0.81 (95% CI 0.70 to 0.94), P value
including one or more of the factors of interest to us (Althausen
= 0.006, I2 = 0%, indicating that for temporal lobe surgery only, 2013; Boesebeck 2007; Cossu 2005; Cossu 2008; Elsharkawy
participants without postoperative discharges are significantly 2008a; Elsharkawy 2009a; Gelinas 2011; Grivas 2006; Janszky
more likely to achieve a good outcome of surgery than those 2003a; Jennum 1993; Kim 2009; Kim 2010a; Lopez-Gonzalez 2012;
with postoperative discharges. However, within the three studies Madhavan 2007; McIntosh 2012; O'Brien 2000; Paolicchi 2000;
in which a mixture of temporal or extratemporal lobe surgeries Phi 2009; Radhakrishnan 1998; Rossi 1994; Sagher 2012; Sarkis
were performed on participants, the pooled RR for presence versus 2012; Schramm 2011; Spencer 2005; Tezer 2008; Theodore 2012;
absence of postoperative discharges is 1.20 (95% CI 0.89 to 1.61), P Walz 2003; Wyler 1995; Yang 2011). In two studies, two separate
value = 0.23, I2 = 81%. This shows a non-significant trend in favour multivariate models were described for seizure outcome with
of the presence of postoperative discharges for a good outcome of different combinations of variables entered into the two models
surgery and all of the heterogeneity in the analysis is between these (McIntosh 2012; O'Brien 2000).
three studies. These three studies are the smallest in this analysis
and all show large, imprecise effect sizes due to small participant We extracted data from all 29 studies regarding variables entered
numbers. Further, the mixture of participants undergoing temporal into the multivariable model, statistical methods used, results of
lobe and extratemporal lobe surgery in these studies is likely to multivariable modelling and we performed an additional risk of
have contributed to the variability. The difference between the bias assessment using a tool for prognostic studies on each of the
subgroups also may suggest a difference between the presence of 29 studies (see Risk of bias in included studies). Overall, we judged
postoperative discharges in participants undergoing temporal or one study to be at high risk of bias in three domains (Rossi 1994),
extratemporal surgery and also the prognosis of the two surgery three studies to be at high risk of bias in two domains (Althausen
types. 2013; Janszky 2003a; Madhavan 2007), 11 studies to be at high risk
Informed decisions.

of bias in one domain (Cossu 2008; Elsharkawy 2008a; Elsharkawy In one study, a multivariable logistic regression analysis is
2009a; Jennum 1993; Kim 2009; Paolicchi 2000; Radhakrishnan described for a wide range of prognostic variables (including the
1998; Sarkis 2012; Spencer 2005; Tezer 2008; Theodore 2012), and extent of surgical resection of interest to us), however reference
we did not judge 13 studies to be at high risk of bias in any domain is made to "univariate conditions" when the results of the model
(Boesebeck 2007; Cossu 2005; Gelinas 2011; Kim 2010a; Lopez- are described (Paolicchi 2000). Therefore it is unclear whether the
Gonzalez 2012; McIntosh 2012; O'Brien 2000; Phi 2009; Sagher 2012; results presented are adjusted or unadjusted.
Schramm 2011; Walz 2003; Wyler 1995; Yang 2011).
Adjusted results
In the 29 studies, 2311 individuals out of 3564 who underwent In the other 26 studies, adjusted results are presented for a
surgery experienced a good outcome of surgery (65%). Five multivariable model. Results are summarised below according to
studies identified participants according to a prospective design the level of detail reported in the results (least detailed to most
(Radhakrishnan 1998; Schramm 2011; Spencer 2005; Theodore detailed) and the type of regression model fitted.
2012; Wyler 1995), of which two had a randomised design
(Schramm 2011; Wyler 1995). In one study it was not clear In four studies, only the significance of results was reported (Cossu
if participants were identified according to a prospective or 2005; Elsharkawy 2008a; Elsharkawy 2009a; Theodore 2012). In all
retrospective design (Walz 2003); the remaining 23 studies four studies it was clear which variables had been entered into the
identified participants according to a retrospective design. model, however in two studies, Elsharkawy 2008a and Elsharkawy
2009a, variables were entered into the multivariable model based
In one study, it was unclear from the information reported in the on statistical significance in univariate analyses and in all four
study whether the dependent variable of seizure outcome was studies "significant" or "non-significant" results were selectively
analysed as 'good' (Engel Class 1) or 'favourable' (Engel Class 1 and reported.
2) (Grivas 2006). Furthermore, it was unclear in this study which
variables were entered into the model and very little information • In Cossu 2005, a large range of variables are entered into
was reported regarding the results of the modelling. Due to these a stepwise multivariable logistic regression analysis: discrete
issues, this study is not considered any further in the narrative numerical variables (age at seizure onset, illness duration before
review of multivariable models. surgery, monthly seizure frequency) and categorical variables
(sex, presence of lesion on MRI, presence of MTS on MRI, side
Given the variability of statistical regression models used (logistic of stereoelectroencephalography (SEEG), site of epileptogenic
regression, Cox proportional hazards regression, generalised zone (EZ), site of resection, type of resection, completeness of
estimating equations etc.), combinations of variables entered into lesionectomy, reason for incomplete lesionectomy, histological
prognostic models and the level of detail reported regarding results diagnosis, presence of MTS at histological analysis on resected
of prognostic models, the combination of adjusted results in mesial temporal specimens). The majority of variables are
meta-analysis was impossible in this review. Instead a narrative not mentioned in the results (including presence of MTS
description of the multivariable models fitted is summarised below. on MRI and presence of MTS at histological analysis of
Furthermore, we identified several issues when considering the interest to us) and therefore assumed to be not significantly
multivariable models reported in 28 studies (minus Grivas 2006, see associated with seizure outcome. It is stated that completeness
'Adjusted results' below for further details): of lesionectomy is "statistically significant association with
• In 13 out of 28 studies (46%), it is not clear exactly which outcome at univariate and multivariate analysis" but no P value,
variables had been entered into the model. treatment effect size or direction of effect is mentioned.
• In 13 out of 28 studies (46%), variables seem to have been • In Elsharkawy 2008a, the following variables (significant in
entered into the model according to statistical significance univariate analyses) are entered into multivariable logistic
rather than clinical relevance (i.e. only variables showing a regression analysis: well-circumscribed lesion on the pre-
statistical association with outcome in univariate analyses were operative MRI scan, short duration of epilepsy, early surgical
considered for multivariable modelling). interference, a neoplasm in the resected specimen, a psychic
aura, versive seizures, tonic-clonic seizures, history of previous
• In 13 out of 28 studies (46%), no adjusted treatment effect
surgery, focal cortical dysplasia in the resected specimen. The
sizes were reported (only P values or whether results were
only variable of interest to us is "presence of focal cortical
"significant" were reported).
dysplasia in the resected specimen, which was "unable to
• In 15 out of 28 studies (54%), results of the multivariate model provide predictive value for the outcome".
were selectively or not fully reported (e.g. numerical results are
• In Elsharkawy 2009a, the following variables (significant in
only reported for variables that were shown to be statistically
univariate analyses) are entered into multivariable stepwise
significant in multivariable analysis).
Cox proportional hazards regression analysis: presence of
In one study, adjusted results are not reported for the Cox regional hippocampal atrophy on pre-operative MR imaging;
proportional hazards model fitted to a wide range of prognostic history of febrile seizure; EEG unilateral seizure onset; age
variables considered in the study (including extent of surgical under 30 years at surgery; exclusively unilateral sharp waves;
resection, history of febrile seizures, history of initial precipitating right-sided resection; short epilepsy duration; family history of
insult, distribution of interictal spikes and side of surgery of interest epilepsy; bilateral sharp waves; seizure onset bilateral; versive
to us); only 'crude' unadjusted hazard ratios are reported from the seizures; somatosensory aura; EEG at six months with interictal
model (Walz 2003). epileptiform discharges (IED); and EEG at two years with IED.
The variables of interest to us (history of febrile seizures, EEG
unilateral seizure onset, EEG with IED at two years and right-
sided resection) "did not retain significance in stepwise model."
Informed decisions.

• In Theodore 2012, three variables were entered into a • In Sarkis 2012, the following variables (significant in univariate
multivariable stepwise logistic regression analysis: MRI, 18F- analyses) are entered into multivariate Cox proportional
FCWAY PET and 18F-FDG PET. The only variable of interest to us hazards regression: the type of resection, pre-operative auras,
(MRI) "was not selected as being predictive of outcome". incomplete resection, the presence of postoperative spikes, age
of seizure onset, prior epilepsy surgery. From the variables of
In six studies, only P values for results were reported (Althausen interest to us: incomplete surgical resection and presence of
2013; Jennum 1993; Madhavan 2007; Rossi 1994; Sarkis 2012; postoperative spikes are "correlated with seizure recurrence" (P
Schramm 2011). In three studies, Madhavan 2007, Rossi 1994 and value = 0.03 and P value = 0.0003, respectively)
Sarkis 2012, variables were entered into the multivariable model • In Schramm 2011, it is unclear which variables from a list
based on statistical significance in univariate analyses and in four of "confounders" were entered into multivariable logistic
studies multivariable model results were selectively reported based regression analysis. It is stated that "Extent of resection and
on statistical significance (Althausen 2013; Jennum 1993; Sarkis surgery type interacted, as did extent of resection and center
2012; Schramm 2011). (p=0.073)", however any association between extent of resection
• In Althausen 2013, it is specified in the methods section that age and seizure outcome is not discussed. It is unclear if any other
at surgery, side of surgery, aetiology (acquired, developmental, variables of interest to us were included in the model.
progressive), pre-operative intelligence, age at epilepsy onset In three studies, P values plus regression coefficients were reported
and duration of epilepsy would be entered into a stepwise (Boesebeck 2007; Kim 2010a; Sagher 2012). None of the studies
logistic regression but from the reported results it appears that seemed to enter variables into the multivariable model based
hemispherectomy technique or completeness of hemispheric on statistical significance, however it was unclear in all three
disconnection have also been entered into the model. From the studies exactly which variables had been entered into the models.
variables of interest to us, neither completeness of hemispheric Regression coefficients could not be converted to treatment effects
disconnection nor side of surgery showed "significant effect on (e.g. odds ratios) in any of the three studies.
postoperative outcome" (P values > 0.1). It is unclear if any other
variables of interest to us were included in the model. • In Boesebeck 2007, it was unclear which "demographic" and
• In Jennum 1993, the following variables are entered into "histological" data had been entered into a backward stepwise
multiple regression analysis: age, sex, duration of epilepsy, multivariable logistic regression. From the variables of interest
MRI: structural lesion (tumour, arteriovenous malformation to us, tumour on histology was significantly associated with a
or hamartoma) versus normal or localised atrophy, extent good outcome (regression coefficient = 1.358, standard error =
of temporal resection, temporal versus extratemporal focus, 0.67, P value = 0.044) while vascular lesions and malformation of
ictal focus completely resected, interictal focus completely cortical development (MCD) on histology were not significantly
resected, presence of postresection spikes in the operative associated with outcome (regression coefficient = 0.40, standard
electrocorticography. From the variables of interest to us, error (SE) = 0.94, P value = NS and regression coefficient = 0.39,
structural lesion on MRI is not reported in the results of standard error = 0.62, P value = non-significant, respectively)
factors strongly associated with good outcome. P values of < • In Kim 2010a, extent of resection, EEG findings, MRI findings
0.01 and < 0.04 are reported for complete resection of ictal and pathology findings (not further defined) were entered into
focus and presence of postresection spikes in the operative multivariate logistic regression analysis. From the variables of
electrocorticography respectively with no further discussion of interest to us: visible lesion on MRI and extent of resection of
the association between these variables and outcome. electrodes were significantly associated with outcome (P value
• In Madhavan 2007, the following variables (significant in < 0.001, likelihood ratio 12.7 and P value = 0.008, likelihood ratio
univariate analyses) are entered into multivariate analysis 7.1 respectively), however the direction of association was not
(details of multivariate model not specified): age at seizure specified. It also appears that presence of tumour and presence
onset; presence/prior history of infantile spasms; ictal and of cortical dysplasia on pathology were entered into the model
interictal focality (unilateral versus bilateral); extent of surgical but no numerical results are mentioned for these variables.
procedure (corpus callosotomy, lesionectomy, lobar resection, • In Sagher 2012, the following variables were analysed via
lesionectomy and lobar resection); presence of residual generalised estimating equations: type of operation, the
dominant tube. From the variables of interest to us: "there was a extent of resection of mesial temporal structures, sex,
trend toward better seizure outcome in those patients that had patient age at operation, handedness, pre-operative secondary
more extensive procedures (p=0.86)" and interictal focality was generalisation, age of onset of epilepsy, the duration of
not associated with seizure outcome (P value = 0.42). a patient's epilepsy, dual pathology, complications, grid
• In Rossi 1994, the following variables (significant in univariate implantation. From the variables of interest to us, increasing
analyses) are entered into multivariate logistic regression: percentage of various mesial temporal structures (amygdala,
spatial arrangement of the electrocerebral epileptiform hippocampus, entorhinal cortex and total) was significantly
interictal and ictal activities, the extent of resection of the associated with good outcome and implantation of grids and/
structural lesion and that of the epileptogenic zone. From the or depth electrodes was significantly associated with poor
variables of interest to us: extent of resection of the structural outcome (coefficient = 0.12, SE = 0.06, P value = 0.05).
lesion and the prevalence of interictal epileptiform activity are
"significantly associated with surgical outcome" (P value < 0.001 In two studies a mixture of P values and treatment effect sizes were
and P value = 0.013 respectively), however the direction of reported (Janszky 2003a; Spencer 2005).
association was not specified for either variable.
• In Janszky 2003a, the following variables were entered into
stepwise logistic regression analyses (variables not selected

Informed decisions.

based on statistical significance): complex febrile seizures of macroscopic total resection was significantly associated with
(CFCs), unilateral interictal epileptiform discharges, tonic–clonic poor outcome: RR 18.22, 95% CI 3.81 to 87.09, P value < 0.001. It
seizures, perinatal insult or side of the surgery. Numerical is unclear if any other variables of interest to us were included
results were reported only for variables which were significant in the model.
in multivariable analysis; from the variables of interest to us:
history of CFC, odds ratio (OR) 5.9, 95% CI (1.26 to 27.7), P value The remaining eight studies reported odds ratios from
= 0.023 where OR > 1 indicates that history of CFC is associated multivariable logistic regression models (Cossu 2008; Gelinas 2011;
with a good surgical outcome. It was also stated that history of Kim 2009; O'Brien 2000; Radhakrishnan 1998; Tezer 2008; Wyler
perinatal insult, unilateral epileptiform discharges concordant 1995; Yang 2011). In four studies variables were entered into the
with the operated side and side of surgery had "no influence on multivariable model based on statistical significance in univariate
post-operative outcome." analyses (Cossu 2008; Kim 2009; Radhakrishnan 1998; Tezer 2008),
• In Spencer 2005, it was unclear which of the variables from in five studies it was not clear exactly which variables had
a long list of "independent variables" were entered into the been entered into multivariable analysis (Gelinas 2011; Kim 2009;
multivariable proportional hazards model; variables did not Radhakrishnan 1998; Tezer 2008; Yang 2011), and in three studies
seem to be entered based on statistical significance. Numerical multivariable model results were selectively reported based on
results were reported only for variables that were significant statistical significance (Cossu 2008; Kim 2009; Tezer 2008).
in multivariable analysis. From the variables of interest to us: • In Cossu 2008, the following variables (significant in univariate
presence of hippocampal atrophy on MRI, RR 1.58 (95% CI 1.13 analysis) were entered into multivariable analysis: sex,
to 2.21), P value = 0.007, where RR > 1 indicates that presence neurologic status, age at seizure onset, duration of epilepsy,
of hippocampal atrophy on MRI is a predictor of two-year seizure frequency, MRI findings, use of Video EEG and SEEG, age
remission. It was also stated that a history of febrile seizures did at surgery, type of surgery, side and site of surgery, extent of
not "approach significance"; it is unclear if any other variables of lesion resection, histology of resected tissue, length of follow-
interest to us were included in the model. up. From the variables of interest to us: complete lesionectomy
Three studies reported risk ratios (RRs) or hazard ratios (HRs) from and histologic diagnosis of neuronal/glial-neuronal tumour
multivariable Cox proportional hazards models (Lopez-Gonzalez were significantly associated with surgical outcome (OR 0.40,
2012; McIntosh 2012; Phi 2009). All three studies entered variables 95% CI 0.23 to 1.01, P value = 0.05 and OR 0.19, 95% CI 0.05 to
into multivariable analysis based on statistical significance at 0.53, P value = 0.004, respectively).
univariate analysis but none of the studies seemed to selectively • In Gelinas 2011, it was unclear which of the variables from
report the results of multivariable analysis based on statistical "clinical characteristics" were entered into analysis. From the
significance. variables of interest to us, use of ECoG (intracranial monitoring)
was not associated with seizure freedom at one year or at most
• In Lopez-Gonzalez 2012, the following variables (significant in recent follow-up (OR 0.71, 95% CI 0.21 to 2.43, P value = 0.59 and
univariate analyses) are entered into multivariable analysis: OR 0.44, 95% CI 0.14 to 1.34, P value = 0.15, respectively). It is
side of surgery (left), number of antiepileptic drug trials (more unclear if any other variables of interest to us were included in
than four), tumour aetiology and extensive surgical resection the model.
(lesionectomy plus cortico-amygdalohippocampectomy). From • In Kim 2009, it was unclear which of the variables from a long
the variables of interest to us: tumour aetiology and right side list of "clinical characteristics" and "prognostic factors" were
of resection were not significant predictors of seizure freedom significant variables at univariate analysis and had been entered
(RR 0.72, 95% CI 0.30 to 1.48 and RR 1.59, 95% CI 0.92 to 2.76, into analysis. From the variables of interest to us: complete
respectively) while extensive surgical resection was a significant resection of epileptogenic area was associated with seizure
predictor of seizure freedom (RR 0.13, 95% CI 0.007 to 0.64, P outcome (OR 4.94, 95% CI 2.41 to 10.14, P value < 0.001). It is
value = 0.007). unclear if any other variables of interest to us were included in
• In McIntosh 2012, two multivariable analyses were conducted. the model.
The first model entered the following variables: pathology • In O'Brien 2000, two multivariable analyses were conducted.
(lesion (tumour), acquired insult, cortical dysplasia, non- The first model entered the following variables: Subtraction ictal
specific) and extent of resection. The second model entered the single-photon emission computed tomography co-registered
variables that were significant in the first multivariable analysis with MRI (SISCOM) regional localisation (concordant versus
plus presence or absence of early postoperative seizures. From non-concordant/non-localising), pre-operative MRI findings
the variables of interest to us, model one shows that incomplete (lesional versus non-lesional) and ictal scalp EEG findings
resection is a significant risk factor for seizure recurrence (HR (localising versus non-localising). The second model entered
1.71, 95% CI 1.06 to 2.76, P value = 0.028) but is no longer a the following variables: extent of excision of the SISCOM focus,
significant risk factor in model two (HR 0.98, 95% CI 0.59 to 1.63), pre-operative MRI findings and ictal scalp EEG findings. Neither
P value = 0.95). Also, in both models one and two pathology of model showed the presence of focal structural lesion to be a
FCD compared to other pathologies is a significant risk factor for significant predictor of surgical outcome (OR 2.3, P value = 0.32
seizure recurrence (model one: HR 1.90, 95% CI 1.08 to 3.34, P and OR 4.7, P value = 0.28, respectively). The second model
value = 0.025, model two: HR 1.54, 95% CI (0.87 to 2.42), P value showed complete excision of SISCOM focus to be a significant
= 0.014). predictor of surgical outcome (OR 201.0, P value = 0.03).
• In Phi 2009, it was unclear which of the variables from a long • In Radhakrishnan 1998, it was unclear which of the variables
list of "patient related factors" and "treatment related factors" were significant variables at univariate analysis and had been
were significant variables at univariate analysis and had been entered into analysis. From the variables of interest to us:
entered into analysis. From the variables of interest of us, failure
Informed decisions.

the presence of MRI-detected unilateral hippocampal formation However, we note that for all but one study (O'Brien 2000), we had
atrophy "is a predictor of excellent seizure control" (OR 3.7, 95% judged the statistical analysis and reporting of the multivariable
CI 1.2 to 11.71, P value = 0.024). It is unclear if any other variables models to be at moderate or high risk of bias and therefore caution
of interest to us were included in the model. is recommended when interpreting any numerical results from
• In Tezer 2008, it was unclear which "continuous" and the multivariable models. Many of the multivariable models only
"categorical" variables were significant variables at univariate included variables that were significant at univariate analysis or
analysis and had been entered into analysis. From the variables in which some variables did not retain significance in adjusted
of interest to us: presence of history of trauma is "predictive for analysis (in other words, the aim of the model may have been
worsening of outcome" (OR 0.33, 95% CI 0.003 to 0.38, P value to investigate which variables would not retain significance). This
= 0.05). It is unclear if any other variables of interest to us were approach to multivariable analysis may explain why no significant
included in the model. association was found between certain variables and outcomes
• In Wyler 1995, the following variables were entered into in some of the studies, whereas a univariate association had
multivariable analysis: extent of resection (partial versus been shown in this review. Furthermore, in many of the studies,
total hippocampectomy), age at surgery, age at onset of results were only reported for variables that showed a statistically
epilepsy, laterality of surgery, gender and presence/absence of significant association with outcome. Therefore there may have
hippocampal sclerosis. From the variables of interest to us: total been more instances where a factor of interest to us was shown not
hippocampectomy (compared to partial hippocampectomy) to be associated with outcome by a multivariable model but this
was associated with seizure freedom (OR 4.2, P value = 0.02), information is not published in the study paper.
but presence of hippocampal sclerosis was not associated with
There were several instances where variables were reported to be
seizure freedom (OR 2.0, P value = 0.32).
associated with outcome (particularly extent of resection) but the
• In Yang 2011, it was unclear which of the variables from "data direction of the association (variably associated with good or poor
collected" were entered into analysis. From the variables of outcome) was not specified; reporting an association without a
interest to us: concordance of MRI with EEG and presence direction is of little clinical utility for practice.
of hippocampal sclerosis were significantly associated with
outcome (OR 6.33, 95% CI 1.44 to 27.81, P value = 0.015 and DISCUSSION
OR 18.06, 95% CI 4.48 to 72.88, P value < 0.001, respectively),
while complete resection and use of subdural electrodes were Summary of main results
not significantly associated with outcome (OR 0.60, 95% CI 0.06
We have identified 177 studies investigating the outcome of surgery
to 5.76, P value = 0.659 and OR 0.61, 95% CI 0.14 to 2.68, P value
for epilepsy. These studies were of variable size and design, were
= 0.52, respectively).
conducted in a range of countries and recruited a wide range
Summary of adjusted analyses of participants of different ages and with different durations of
epilepsy. A wide range of surgical techniques were carried out
It is difficult to summarise the adjusted results of the 26 across these studies and different scales were used to measure the
studies described above collectively due to the different statistical outcome of surgery.
methods used, the different variables entered into the models
chosen according to different criteria and the variable levels of Four of the 177 studies had a randomised controlled trial (RCT)
detail in the reporting of multivariable results. design: one that randomised surgery and medical treatment, two
that randomised types of surgical technique (anterior temporal
Overall, none of the reported prognostic models included lobectomy (ATL) with or without corpus callosotomy and partial
encephalomalacia as a factor, one included vascular or total hippocampectomy respectively) and one that randomised
malformations, two included concordance of MRI and EEG, length of surgical resection (2.5 cm versus 3.5 cm resection).
head injury and distribution of interictal spikes, three included
focal cortical dysplasia, febrile seizures and use of intracranial The RCT that randomised surgery and medical treatment found
monitoring, four included MRI results, tumour and postoperative surgery to be superior to medical treatment in terms of freedom
discharges, five included side of resection, six included mesial from all seizures including auras at one year (risk ratio (RR)
temporal sclerosis and 18 included extent of resection. 15.00, 95% CI 2.08 to 108.23, P value = 0.007). There was no
statistically significant difference between ATL with or without
Due to the different combination of variables entered into corpus callosotomy in terms of seizure freedom at two years (RR
multivariable models and therefore the adjusted results not being 1.22, 95% CI 0.85 to 1.76, P value = 0.28), or between 2.5 cm or 3.5 cm
adjusted for the same factors, direct comparison of the multivariate ATL resection in terms of seizure freedom at one year (RR 1.02, 95%
analysis with the univariate meta-analysis reported in this review is CI 0.86 to 1.20, P value = 0.84), respectively. There was a statistically
difficult. significant advantage of total over partial hippocampectomy in
terms of seizure freedom at one year (RR 1.82, 95% CI 1.12 to 2.93,
Generally the direction of the association of variables with outcome
P value = 0.01). We judged the evidence from the four RCTs to be of
was the same in the multivariable analyses to that in the
moderate to low quality according to the GRADE criteria due to the
univariate meta-analyses. The only differences of note were that no
lack of reported information about randomised trial design and the
multivariable models found an association between side of surgery
restrictive study populations (three out of the four RCTs included
and outcome (the significant advantage for right-sided resection
only adults and excluded children).
in Analysis 4.14) or between distribution of interictal spikes and
outcome (the significant advantage for unilateral spikes in Analysis There were 16,253 participants included in the 177 studies in this
4.11). review of which 10,518 (65%) achieved a good outcome of surgery,

Informed decisions.

ranging across studies from 13.5% to 92.5% of participants. We unlikely that any evidence not identified by the electronic searches
found the reporting of related adverse events to be sparse and very would change the review conclusions.
poor; complications and/or surgery-related deaths were reported
in less than half of included studies, often lacking specific detail of We were able to extract data for 16,508 participants from 177
the nature and consequence of the event (transient or permanent) eligible studies and the large number of participants we were able
and the timing of events. Few studies contained any reference to to include in the review is reflected in the precision of our results.
postoperative cognition or mental state. However, we believe that some of the studies we had to exclude
from the review (see Excluded studies) could possibly have been
One hundred and eighteen studies examined between one and nine included if additional information regarding follow-up had been
factors of interest in this review in univariate relation to outcome available or if data had been presented in an extractable way.
of surgery. We found abnormal pre-operative magnetic resonance For example, a number of studies reported only percentages, P
imaging (MRI), no use of intracranial monitoring, complete surgical values or coefficients of regression models for prognostic factors
resection, presence of mesial temporal sclerosis, concordance of related to outcome and we were unable to extract raw data for
pre-operative MRI and electroencephalography (EEG), history of the number of participants with and without the prognostic factor
febrile seizures, absence of focal cortical dysplasia/malformation achieving good or poor outcome of surgery for inclusion in our
of cortical development, presence of tumour, right-sided resection univariate analyses and insufficient detail regarding multivariable
and presence of unilateral interictal spikes to be independently analyses of prognostic factors was reported to allow for meaningful
associated with a good outcome of surgery. We found no evidence comparison with univariate analyses.
that history of head injury, presence of encephalomalacia, presence
of vascular malformation or presence of postoperative discharges Furthermore, several studies reported outcomes as
were independent, univariate prognostic factors for outcome of 'favourable' (usually Engel Class 1 and 2 combined) versus
surgery. We observed variability between studies for many of our 'unfavourable' (Engel Class 3 and 4 combined) rather than good
analyses, likely due to small study sizes with unbalanced group versus poor outcome (Engel Class 1 versus Engel Classes 2 to 4).
sizes, variation in the definition of the seizure outcome, definition We strongly believe that there is a large difference between a
of the prognostic factor (for example, the definition of a 'complete' 'good' outcome and a 'favourable' outcome of surgery for epilepsy.
resection varied across studies) and the influence of the site of Therefore we did not feel it was appropriate to extract data on
surgery, which we have observed to be related to postoperative prognostic factors expressed as 'favourable' versus 'unfavourable.'
seizure outcome. However, if the data had been presented in a different way, for
example each Engel Class 1 to 4 by the prognostic factor of interest,
Twenty-nine studies presented multivariable prognostic models then we would have been able to include this information, further
of which 26 presented clear adjusted results for the association improve the precision of our results and possibly reduce variability
of independent variables with the dependent variable of seizure between studies in our results.
freedom. None of the studies included encephalomalacia as
a factor, one included vascular malformations, two included Applicability of our results
concordance of MRI and EEG, head injury and distribution of The potential efficacy of temporal and extratemporal resection
interictal spikes, three included focal cortical dysplasia, febrile for people with a focal epilepsy uncontrolled by antiepileptic
seizures and use of intracranial monitoring, four included MRI medication is undisputed (Engel 2003) and has been confirmed by
results, tumour and postoperative discharges, five included side of our review, which showed that about two-thirds of patients have
resection, six included mesial temporal sclerosis and 18 included a good surgical outcome. For temporal resections Engel's practice
extent of resection. Generally the direction of the association parameter focuses on the single intention-to-treat randomised
of variables with outcome was the same in multivariable controlled trial of surgery for mesial temporal lobe epilepsy
analyses to that in univariate meta-analyses. However, due to (Wiebe 2001), which found that 58% of patients randomised to be
different combinations of variables entered into the multivariable evaluated for surgical therapy (64% of those who received surgery)
models, different (often inappropriate) statistical approaches were free of disabling seizures at one year, compared with 8%
used to modelling and selective reporting of results, meaningful free of disabling seizures in the group randomised to continued
comparison of multivariate analysis with the univariate meta- medical therapy. The recent consensus from the International
analysis reported in this review is difficult. Caution is recommended League Against Epilepsy, Kwan 2010b, proposes that a treatment's
when interpreting the results of the univariate meta-analysis and success should be defined by sustained freedom from seizures,
the narratively reported multivariable analyses. as that is the only efficacy outcome consistently associated with
improved quality of life (and in the UK the only efficacy outcome
Overall completeness and applicability of evidence that allows a patient to drive legally). This justifies our focus on
Completeness of evidence at least 12 months' seizure freedom in this review. Using this
measure the meta-analysis by Costa 2011 showed that the overall
We were aware that the majority of the evidence for this review weighted pooled risk difference in favour of the newer antiepileptic
was likely to be from non-randomised studies, however there is drugs compared with placebo for freedom from seizures during the
no Cochrane approved or recommended search filter for non- limited study periods was only 6% (95% confidence interval 4% to
randomised studies. Search strategies were therefore chosen by 8%; number needed to treat in terms of freedom from seizures with
the Cochrane Epilepsy Group's Trial Search Co-ordinators and the newer drugs as add-on therapy ranged from 9 to 19, mean 11.3). The
resulting search filters were not subjected to any systematic testing message for the clinician in the consultation facing a person with
before use. We do believe that our systematic electronic searches intractable epilepsy is therefore clear. There is a one in 11 chance
identified the vast majority of relevant evidence for this review of helping with the next antiepileptic drug and a two out of three
and given the large number of studies included in this review, it is chance of helping with surgery if selection criteria are met.
Informed decisions.

Furthermore, we have observed that overall the surgical lesions often have a more diffuse pathology, many of which are
procedures have a low rate of complications and this is in less easily resectable resulting in a poor surgical outcome. In this
line with other reports showing that less than 5% of patients heterogeneous group we observed inferior 12-month remission
have permanent postoperative neurological deficits secondary to pooled risk ratios: encephalomalacia RR for present versus absent:
accidental damage of central nervous system (CNS) tissue (Engel 0.67, 95% CI 0.37 to 1.21; neuronal migration disorders including
1996; Engel 2003). As stated, the figure given here for adverse events focal cortical dysplasias RR for present versus absent: 0.90, 95%
of 7.3% is highly likely to be an over-estimate as a prevalence CI 0.85 to 0.95 and vascular malformations RR for present versus
figure for permanent neurological deficit as many studies did not absent: 1.07, 95% CI 0.94 to 1.21. In these instances the margin both
record which events were only transient and more than one event on the MRI scan and at operation is often difficult to define.
could be recorded in the same person. Having said that, very few
studies address the important issue of formally reassessing any The good outcome associated with the presence of any abnormal
postoperative impairment of cognition, speech and language and MRI findings is expected and probably largely reflects a variety
social functioning or altered mental state (all linked to quality of of CNS conditions, which we know are associated with a good
life). This aspect should be written into research protocols. prognosis, such as tumours, mesial temporal sclerosis (MTS) and
many congenital malformations. These are discrete structural
The poor recording of adverse event data in the studies is lesions that lend themselves to complete resection. We note
lamentable. The majority of studies recorded no data at all. There moderate heterogeneity between studies in part accounted for by
is no significant improvement seen over the past 10 years since the degree of definition of MRI abnormality in individual studies.
the Tonini 2004 review. When any intervention study is carried out So, non-specific or ill-defined white matter abnormalities in one
there ought to be detailed assessments of both risks and benefits. study may well dilute the benefit of seeing a well-defined area of
We are surprised how journal editors continue to accept papers MTS or a tumour in another study. A reasonable prospect of seizure
for publication without adverse event recording. The required freedom is not ruled out with a normal MRI. Bell 2009 reported an
standard should be to record how many events are recorded in how Engel 1 outcome at 12 months in 60% (24 of 40) people following
many participants and to make it clear which are postoperative anterior temporal lobectomy for medically refractory temporal
complications, which are transient events (within a set period) lobe epilepsy; but it should be noted that one of the inclusion
and which are permanent new impairments. Protocols should criteria was subtle non-specific MRI findings in the mesial temporal
include pre- and postoperative measures of speech and language lobe concordant to the area of resection. Jayakar 2008 reported on
function, cognition and social functioning along with a mental state a cohort (predominantly of children) with non-lesional intractable
assessment. focal epilepsy undergoing resective surgery. At two-year follow-up,
44 of 101 participants were seizure-free. Outcomes correlated with
The criteria adopted to identify the indications and the applications good outcome were the presence of convergent scalp EEG focal
of epilepsy surgery are constantly evolving. Continuing attempts interictal spikes (P value < 0.005) and completeness of resection (P
need to be made to define patient and procedure-related value < 0.0005). Dorward 2011 studied children with extratemporal,
prognostic indicators. Our systematic review shows that the non-lesional epilepsy. Outcome was classified as Engel class 1 or 2
strongest predictors of success of surgery include, in decreasing in 54.5% of the children who underwent resection of the lesion or
order, the extent of resection (especially for extratemporal surgery), multiple sub-pial resections. The results were obtained through the
an abnormal MRI finding, concordance between neuroimaging use of invasive monitoring with grid/strip electrodes.
and EEG findings, tumours, mesial temporal sclerosis, unilateral
EEG discharges especially when extratemporal, a history of There is a strong association between MTS and a history of febrile
febrile seizures and a vascular malformation. By contrast, adverse seizures. The term mesial temporal sclerosis as an alternative
prognostic factors include the need for intracranial monitoring and to hippocampal sclerosis was introduced in recognition of the
the presence of postoperative discharges (particularly those seen frequent involvement of mesial limbic structures adjacent to the
extratemporally). hippocampus. Neocortical neuronal loss and gliosis (temporal lobe
sclerosis (TLS)) has been studied by Thom 2009. They identified
The extent of resection was the strongest determinant of outcome TLS in 30 of 272 surgically treated cases of hippocampal sclerosis.
in our review, particularly with extratemporal surgery. The dilemma There was a history of a febrile seizure as an initial precipitating
for the surgeon is that larger areas of resection are likely to be injury in 73% of patients with TLS compared with 36% without TLS.
associated with a higher complication risk. The extent of resection This was febrile status in 27% of these. The changes of TLS may be
is affected by the underlying pathology, the site of surgery and due to an enhanced vulnerability of superficial cortical neurons to
the development of investigational and surgical procedures. It is an early cerebral event in the maturing neocortex in a small group
clear that this is one area where further randomised controlled of children. The good outcome associated with a history of febrile
trials may inform future practice with good effect. We identified seizures can be interpreted in the light of its association with mesial
two trials with this design (Schramm 2011 (and Helmstaedter 2011 temporal sclerosis.
reporting on a sub-set from Schramm 2011) and Wyler 1995). The
primary intention-to-treat analysis did not show a benefit in seizure EEG/MRI concordance is correlated with positive surgical outcome.
freedom rate for the more extensive resection group in Schramm It is clear that results obtained from individual studies will depend
2011 but did demonstrate benefit in Wyler 1995. very much on the mix of associated pathologies underlying the
epilepsy seen in the participants. Studies containing a large
Tumours carry a higher chance of seizure remission at 12 months number with discrete lesions such as tumours are likely to show
than other CNS disorders (RR present versus absent for 12- more concordance and a better outcome than studies with a
month remission: 1.23, 95% CI 1.14 to1.32). This finding is readily predominance of less discrete lesions, as can be the case with many
understood as the epileptogenic area is more easily detectable and neurodevelopmental abnormalities.
defined both on neuroimaging and at operation. Non-tumoural
Informed decisions.

The need for intracranial monitoring itself implies that there unanswered today by this large body of literature as they did when
is uncertainty in the location and extent of an epileptogenic Victor Horsley helped the young Scot in 1886.
zone, often accompanied by indeterminate neuroimaging. The
association between the need for neuroimaging and a poor There are many questions to answer but they should address
outcome is therefore not surprising. In these cases, fewer than the issues of extent of resection for temporal and extratemporal
50% of cases may become seizure-free postoperatively. Poor lesions, the definition of care pathways for the most cost-efficient
localisation is also reflected in the fact that participants with and effective pre-operative selection, non-lesional focal epilepsy,
unilateral interictal spikes are significantly more likely to achieve a bilateral and postoperative spikes, and when to tail antiepileptic
good outcome of surgery than participants with bilateral interictal drugs, among many others. Clear data on risks (adverse events,
spikes. The persistence of postoperative discharges is likely to their nature and timing) as well as benefits should always
reflect these very same issues. There are units who will re-operate be recorded. Protocols should include pre- and postoperative
very quickly when postoperative discharges are identified. The measures of speech and language function, cognition and social
heterogeneity of our results does not allow us to support this functioning along with a mental state assessment. There will be
approach. It is but one example of how properly conducted methodological difficulties and adequate powering will require
research should in the future inform the correct care pathway (see multi-centre approaches. Improvements in the development of
below). cancer care over the past three to four decades have been achieved
by answering well-defined questions through the conduct of
We have already referred to the limitations of this review. These can focused RCTs in a step-wise fashion. The same approach to surgery
be summarised as including: different criteria for seizure outcome; for epilepsy is required.
the variable length of the follow-up; the forced dichotomisation of
each putative prognostic predictor; retrospective designs for the Quality of the evidence
majority of studies that enhance the risk of bias in data collection
Most of the evidence included in this review was of a non-
and presentation and variables that were examined mostly in
randomised, retrospective design. As detailed in Assessment of risk
univariate analyses without considering the role combined effects
of bias in included studies and Risk of bias in included studies,
of prognostic factors or of other (known or unknown) confounders,
at the time of the initiation of this review, we did not know of a
which may be the true prognostic predictors. We minimised
tool recommended by The Cochrane Collaboration to assess the
bias in part by using restrictive criteria for study inclusion and
quality of studies of this design. Following extensive review of
measuring the heterogeneity of the study results. The predictive
the literature relating to quality assessment tools for studies with
value of prognostic factors was usually higher where study result
non-randomised designs, we selected a tool we believed to be
heterogeneity was lower.
appropriate a priori, however we discovered it was not effective at
Despite these limitations, our results provide some clinical separating differences in quality between studies, several criteria
guidance for the selection of the best surgical candidates. Engel were not applicable for most of the studies we identified and the
2003 identified major methodological deficiencies in the published overall quality rating was dictated by one or two criteria.
studies on epilepsy surgery, which included the retrospective
Currently, the quality of evidence for outcomes addressing the
design, the scarcity of data on pre-operative seizures and the
primary objective of the review is of moderate to very low quality.
absence of blinding in seizure outcome assessment. We have to
We downgraded the evidence due to indirectness, risk of bias and
report no quality improvement in the vast majority of the body
imprecision arising from the small sample sizes in the studies
of literature reviewed here, published since Engel's comments
(Summary of findings for the main comparison; Summary of
11 years ago. We would strongly advocate such an improvement
findings 2). The majority of relevant evidence for this review
and the identification of better standards for the assessment of
addresses the secondary objective of the review but comes from
surgical outcome in future studies. We would emphasise the need
studies with a retrospective design and has not enabled us to
for a prospective design. There are examples of this: the studies
explore the effect of multiple prognostic factors in one analysis.
of Helmstaedter 2011, Schramm 2011 and Wyler 1995 already
mentioned who studied the extent of resection and outcome Surgical treatment for epilepsy is the only treatment option for
of temporal lobe surgery; Liang 2010 who reported benefit in a substantial proportion of individuals. Designing randomised
improved quality of life and performance IQ for anterior temporal trials in which participants could be randomised away from an
lobectomy combined with anterior corpus callosotomy in people effective treatment could be considered unethical (the only RCT to
with temporal lobe epilepsy and mental retardation; Oertel 2005 randomise participants to surgical or medical treatment showed a
who demonstrated that a waterjet dissector enables a significant large advantage for surgery over antiepileptic drugs, Wiebe 2001).
reduction of intraoperative blood loss in epilepsy surgery; Velasco Therefore, identifying prognostic factors that are associated with a
2011 who showed that ictal-SPECT did not add localising value good outcome of surgery is a very important research question.
beyond that provided by EEG-video telemetry and structural
MRI that altered the surgical decision and outcome for MTLE- Until higher quality evidence exists, ideally from studies of a
HS patients; and then there is the sentinel work of Wiebe 2001 prospective cohort design that aim to examine differences in
already referred to. For the future the primary outcome measure surgical approaches and specific prognostic factors of interest, it
for intervention studies ought to be seizure freedom at set time is important that we accurately judge the differences in quality of
points with a minimum of one year of follow-up. Assessment the evidence that currently exists. We established from this review
should be blinded and linked to quality of life measurement. that the EPHPP Quality Assessment Tool was not appropriate
The design should be a randomised controlled trial, appropriately for assessing the quality of retrospective studies of surgery.
powered with a focus on specific research questions that remain as After consideration of the existing tools developed for risk of
bias assessment of studies of a non-randomised design listed in

Informed decisions.

chapter 13 of the Cochrane Handbook for Systematic Reviews of operative investigation is probably not required. When one of
Interventions (Higgins 2011) and the ACROBAT-NRSI tool, recently these pointers is absent more sophisticated imaging and EEG
developed by Methods Groups of The Cochrane Collaboration study (which may include intracranial electrodes) is needed. A pre-
(ACROBAT 2014), we were not able to identify any tool which operative assessment of memory function should be carried out
would be appropriate for the majority of studies of the design in on all candidates for temporal lobe surgery. The technology must
this review. Therefore it is essential that an appropriate tool is be used in the setting of a good inter-disciplinary team. Pre- and
developed for the assessment of non-randomised studies of all postoperative assessments should include cognition and mental
designs, including those without comparator control groups. state; neuropsychological and psychiatric evaluations are essential
parts of presurgical evaluation, and should also be scheduled after
Potential biases in the review process surgery, at a minimum at three to four months and one year.
As we have stated, the use of a retrospective design in the
Implications for research
vast majority of the body of literature reviewed allows great
scope for bias in study results with a lack of standardisation in The case is already made for surgical resection of the epileptogenic
data collection, differing durations of follow-up and a lack of zone in intractable focal epilepsy in carefully selected cases.
blinding for seizure assessment, as reflected by the significant Future research should have a prospective cohort or randomised
heterogeneity of the study results. We have also discussed at controlled trial (RCT) design, be appropriately powered and focus
length in Overall completeness and applicability of evidence the on specific issues related to diagnostic tools, the site-specific
limitations associated with the search strategy employed, the surgical approach and other issues such as extent of resection.
quality assessment tool used and the restricted analyses we were Prognostic factors related to the outcome of surgery should
able to perform, particularly in terms of multivariable analyses. be investigated via multivariable statistical regression modelling
where variables are selected for modelling according to clinical
Agreements and disagreements with other studies or relevance and all numerical results of prognostic models are
reviews fully reported. Protocols should include pre- and postoperative
measures of speech and language function, cognition and social
Our results broadly concur with those presented in the review by
functioning along with a mental state assessment. Adverse events
Tonini 2004 (of which RN was a co-author). We have reviewed
must be recorded; journal editors should not accept papers
a substantially larger body of literature. Often this had led to
where adverse events from a medical intervention are not
the identification of greater heterogeneity for studies related to
recorded.The Effective Public Health Practice Project (EPHPP)
specific ocutome predictors. Nonetheless the general implications
Quality Assessment Tool was not appropriate for assessing the
for clinical practice remain the same. This observation serves to
quality of retrospective studies of surgery and to the best of the
emphasise the need for the clinical questions to be refined and
authors knowledge an appropriate tool does not exist and needs
for the emergence of a new body of research with prospective and
to be developed. Improvements in the development of cancer
randomised controlled design so that future clinical practice can be
care over the past three to four decades have been achieved by
better informed by an improved quality of evidence.
answering well-defined questions through the conduct of focused
AUTHORS' CONCLUSIONS RCTs in a step-wise fashion. The same approach to surgery for
epilepsy is required.
Implications for practice
ACKNOWLEDGEMENTS
The poor quality of the data presented in the majority of
the body of literature reviewed, for example due to lack of We are grateful to Cochrane Epilepsy Group Trial Search Co-
uniformity over definitions of outcomes, prognostic factors and ordinators Alison Beamond and Graham Chan for performing all
measurement times, variable populations, retrospective designs electronic searches. We would also wish to thank Dr Stefan Spinty
and inadequate reporting of analysis results, mean that our for help with the German paper, Miss Cerian Jackson for support in
results provide limited clinical guidance for the selection of the post hoc analyses, Mrs Rachael Kelly and Professor Tony Marson,
best surgical candidates. Assessment for surgical selection should Managing Editor and Co-ordinating Editor of the Cochrane Epilepsy
be offered to all people with a focal epilepsy where the first Group (respectively) for comments and support throughout the
two antiepileptic drugs have failed and assessment for surgery review process and the Cochrane Prognostic Methods Group
must be tailored to the individual, considering co-morbidities (particularly Professor Karel Moons) for advice on systematic
and the whole patient context. Considering the results of the reviews of prognostic factors. We should also like to acknowledge
univariate analysis conducted in this review and supported by the the authors of the Tonini 2004 review (of which RN was one) and
multivariable analyses conducted in the included studies, where a in particular to acknowledge the help and generosity of Prof Ettore
discrete lesion on magnetic resonance imaging (MRI) is identified Beghi and Clara Tonini for providing their data and allowing us to
and there is good concordance with seizure semiology and ictal include them in this updated review.
electroencephalography (EEG) discharges more sophisticated pre-

Informed decisions.

REFERENCES

References to studies included in this review Bautista 2003 {published data only}
Aaberg 2012 {published data only} Bautista JF, Foldvary-Schaefer N, Bingaman WE, Luders HO.
Focal cortical dysplasia and intractable epilepsy in adults:
Aaberg KM, Eriksson AS, Ramm-Pettersen J, Nakken KO. Long-
clinical, EEG, imaging, and surgical features. Epilepsy Research
term outcome of resective epilepsy surgery in Norwegian
2003;55(1-2):131-6.
children. Acta Paediatrica 2012;101(12):e557-60.
Bell 2009 {published data only}
Adam 1996 {published data only}
Bell ML, Rao S, So EL, Trenerry M, Kazemi N, Stead SM, et al.
Adam C, Clemenceau S, Semah F, Hasboun D, Samson S,
Epilepsy surgery outcomes in temporal lobe epilepsy with a
Aboujaoude N, et al. Variability of presentation in medial
normal MRI. Epilepsia 2009;50(9):2053-60.
temporal lobe epilepsy: a study of 30 operated cases. Acta
Neurologica Scandinavica 1996;94:1-11. Benifla 2006 {published data only}
Adelson 1992 {published data only} Benifla M, Otsubo H, Ochi A, Weiss SK, Donner EJ, Shroff M,
et al. Temporal lobe surgery for intractable epilepsy in
Adelson PD, Peacock WJ, Chugani HT, Comair YG, Vinters HV,
children: an analysis of outcomes in 126 children. Neurosurgery
Shields WD, et al. Temporal and extended temporal resections
2006;59(6):1203-13.
for the treatment of intractable seizures in early childhood.
Pediatric Neurosurgery 1992;18:169-78. Berkovic 1995 {published data only}
Alfstad 2011 {published data only} Berkovic SF, McIntosh AM, Kalnins RM, Jackson GD, Fabinyi GC,
Brazenor GA, et al. Preoperative MRI predicts outcome
Alfstad KA, Lossius MI, Roste GK, Mowinckel P, Scheie, D,
of temporal lobectomy: an actuarial analysis. Neurology
Borota OC, et al. Acute postoperative seizures after epilepsy
1995;45:1358-63.
surgery - a long-term outcome predictor?. Acta Neurologica
Scandinavica 2011;123(1):48-53. Blount 2004 {published data only}
Althausen 2013 {published data only} Blount JP, Langburt W, Otsubo H, Chitoku S, Ochi A, Weiss S,
et al. Multiple subpial transections in the treatment of
Althausen A, Gleissner U, Hoppe C, Sassen R, Buddewig S,
pediatric epilepsy. Journal of Neurosurgery 2004;100(2 Suppl
von Lehe M, et al. Long-term outcome of hemispheric surgery
Pediatrics):118-24.
at different ages in 61 epilepsy patients. Journal of Neurology,
Neurosurgery & Psychiatry 2013;84(5):529-36. Blume 2004 {published data only}
Arruda 1996 {published data only} Blume WT, Ganapathy GR, Munoz D, Lee DH. Indices of resective
surgery effectiveness for intractable nonlesional focal epilepsy.
Arruda F, Cendes F, Andermann F, Dubeau F, Villemure JG,
Epilepsia 2004;45(1):46-53.
Jones-Gotman M, et al. Mesial atrophy and outcome after
amygdalohippocampectomy or temporal lobe removal. Annals Boesebeck 2007 {published data only}
of Neurology 1996;40:446-50.
Boesebeck F, Janszky J, Kellinghaus C, May T, Ebner A.
Awad 1991 {published data only} Presurgical seizure frequency and tumoral etiology predict
the outcome after extratemporal epilepsy surgery. Journal of
Awad IA, Rosenfeld J, Ahl J, Hahn JF, Luders H. Intractable
Neurology 2007;254(8):996-9.
epilepsy and structural lesions of the brain: mapping, resection
strategies, and seizure outcome. Epilepsia 1991;32:179-86. Boshuisen 2010 {published data only}
Babini 2013 {published data only} Boshuisen K, van Schooneveld MM, Leijten FS, de Kort GA,
van Rijen PC, Gosselaar PH, et al. Contralateral MRI
Babini M, Giulioni M, Galassi E, Marucci G, Martinoni M,
abnormalities affect seizure and cognitive outcome after
Rubboli G, et al. Seizure outcome of surgical treatment of focal
hemispherectomy. Neurology 2010;75(18):1623-30.
epilepsy associated with low-grade tumors in children. Journal
of Neurosurgery. Pediatrics 2013;11(2):214-23. Brainer-Lima 1996 {published data only}
Battaglia 2006 {published data only} Brainer-Lima PT, Rao S, Cukiert A, Yacubian EM, Gronich G,
Marino Jr. R. Surgical treatment of refractory epilepsy
Battaglia D, Chieffo D, Lettori D, Perrino F, Di Rocco C,
associated with space occupying lesions. Experience and
Guzzetta F. Cognitive assessment in epilepsy surgery of
review. Arquivos de Neuro-Psiquiatria 1996;54:384-92.
children. Childs Nervous System 2006;22(8):744-59.
Britton 1994 {published data only}
Baumann 2007 {published data only}
Britton J, Cascino GD, Sharbrough FW, Kelly PJ. Low-grade glial
Baumann CR, Acciarri N, Bertalanffy H, Devinsky O, Elger CE, Lo
neoplasms and intractable partial epilepsy: efficacy of surgical
Russo G, et al. Seizure outcome after resection of supratentorial
treatment. Epilepsia 1994;35:1130-5.
cavernous malformations: a study of 168 patients. Epilepsia
2007;48(3):559-63.

Informed decisions.

Caraballo 2011 {published data only} be independent of age. Clinical Neurology & Neurosurgery
Caraballo R, Bartuluchi M, Cersósimo R, Soraru A, Pomata H. 2009;111(3):240-5.
Hemispherectomy in pediatric patients with epilepsy: a study of
Cukiert 2002 {published data only}
45 cases with special emphasis on epileptic syndromes. Child's
Nervous System 2011;27:2131-6. Cukiert A, Buratini JA, Machado E, Sousa A, Vieira J,
Forster C, et al. Seizure-related outcome after
Cascino 1995 {published data only} corticoamygdalohippocampectomy in patients with refractory
Cascino GD, Trenerry MR, Jack Jr. CR, Dodick D, Sharbrough FW, temporal lobe epilepsy and mesial temporal sclerosis evaluated
So EL, et al. Electrocorticography and temporal lobe epilepsy: by magnetic resonance imaging alone. Neurosurgical Focus
relationship to quantitative MRI and operative outcome. 2002;13(4):ecp2.
Epilepsia 1995;36(7):692-6.
Dagar 2011 {published data only}
Chabardes 2005 {published data only} Dagar A, Chandra P, Gaikwad S, Garg A, Rekha D, Gulati S, et al.
Chabardes S, Kahane P, Minotti L, Tassi L, Grand S, Hoffmann D, Epilepsy surgery in a pediatric population: a retrospective study
et al. The temporopolar cortex plays a pivotal role in temporal of 129 children from a tertiary care hospital in a developing
lobe seizures. Brain 2005;128(8):1818-31. country along with assessment of quality of life. Pediatric
Neurosurgery 2011;47:186-93.
Chang 2009 {published data only}
Dalmagro 2005 {published data only}
Chang EF, Gabriel RA, Potts MB, Garcia PA, Barbaro NM,
Lawton MT. Seizure characteristics and control after Dalmagro CL, Bianchin MM, Velasco TR, Alexandre V Jr, Walz R,
microsurgical resection of supratentorial cerebral cavernous Terra-Bustamante VC, et al. Clinical features of patients with
malformations. Neurosurgery 2009;65(1):31-7. posterior cortex epilepsies and predictors of surgical outcome.
Epilepsia 2005;46(9):1442-9.
Chee 1993 {published data only}
Delbeke 1996 {published data only}
Chee MW, Morris III HH, Antar MA, Van Ness PC, Dinner DS,
Rehm P, et al. Presurgical evaluation of temporal lobe epilepsy Delbeke D, Lawrence SK, Abou-Khalil BW, Blumenkopf B,
using interictal temporal spikes and positron emission Kessler RM. Postsurgical outcome of patients with uncontrolled
tomography. Archives of Neurology 1993;50:45-8. complex partial seizures and temporal lobe hypometabolism on
18FDG-positron emission tomography. Investigative Radiology
Chkhenkeli 2007 {published data only} 1996;31:261-6.
Chkhenkeli S, Towle VL, Lortkipanidze GS, Spire JP,
Dellabadia 2002 {published data only}
Bregvadze ES, Hunter JD, et al. Mutually suppressive
interrelations of symmetric epileptic foci in bitemporal DellaBadia J Jr, Bell W, Keyes JW Jr, Mathews VP, Glazier SS.
epilepsy and their inhibitory stimulation. Clinical Neurology & Assessment and cost comparison of sleep-deprived EEG, MRI
Neurosurgery 2007;109(1):7-22. and PET in the prediction of surgical treatment for epilepsy.
Seizure 2002;11(5):303-9.
Choi 2004a {published data only}
de Tisi 2011 {published data only}
Choi JY, Chang JW, Park YG, Kim TS, Lee BI, Chung SS. A
retrospective study of the clinical outcomes and significant de Tisi J, Bell GS, Peacock JL, McEvoy AW, Harkness WF,
variables in the surgical treatment of temporal lobe tumor Sander JW, et al. The long-term outcome of adult epilepsy
associated with intractable seizures. Stereotactic Functional surgery, patterns of seizure remission, and relapse: a cohort
Neurosurgery 2004;82(1):35-42. study. Lancet 2011;378(9800):1388-95.
Chung 2005 {published data only} Devlin 2003 {published data only}
Chung CK, Lee SK, Kim KJ. Surgical outcome of epilepsy caused Devlin AM, Cross JH, Harkness W, Chong WK, Harding B, Vargha-
by cortical dysplasia. Epilepsia 2005;46(Suppl 1):25-9. Khadem F, et al. Clinical outcomes of hemispherectomy
for epilepsy in childhood and adolescence. Brain
Cossu 2005 {published data only} 2003;126(3):556-66.
Cossu M, Cardinale F, Castana L, Citterio A, Francione S,
Donadio 2011 {published data only}
Tassi L, et al. Stereoelectroencephalography in the presurgical
evaluation of focal epilepsy: a retrospective analysis of 215 Donadio M, D'Giarno C, Moussalli M, Barrios L, Ugarnes G,
procedures. Neurosurgery 2005;57(4):706-18. Segalovich M, et al. Epilepsy surgery in Argentina: long-
term results in a comprehensive epilepsy centre. Seizure
Cossu 2008 {published data only} 2011;20:442-5.
Cossu M, Lo Russo G, Francione S, Mai R, Nobili L, Sartori I, et al.
Dorward 2011 {published data only}
Epilepsy surgery in children: results and predictors of outcome
on seizures. Epilepsia 2008;49(1):65-72. Dorward IG, Titus JB, Limbrick DD, Johnston JM, Bertrand ME,
Smyth MD. Extratemporal, nonlesional epilepsy in children:
Costello 2009 {published data only} postsurgical clinical and neurocognitive outcomes. Journal of
Costello DJ, Shields DC, Cash SS, Eskandar EN, Cosgrove GR, Neurosurgery 2011;7(2):179-88.
Cole AJ. Consideration of epilepsy surgery in adults should
Informed decisions.

Duchowny 1998 {published data only} Garcia 1991 {published data only}
Duchowny M, Jayakar P, Resnick T, Harvey AS, Alvarez L, Dean P, Garcia PA, Barbaro NM, Laxer KD. The prognostic value of
et al. Epilepsy surgery in the three years of life. Epilepsia postoperative seizures following epilepsy surgery. Neurology
1998;39:737-43. 1991;41:1511-2.
Dunkley 2011 {published data only} Garcia 1994 {published data only}
Dunkley C, Kung J, Scott RC, Nicolaides P, Neville B, Aylett SE, et Garcia PA, Laxer KD, Barbaro NM, Dillon WP. Prognostic value
al. Epilepsy surgery in children under 3 years. Epilepsy Research of qualitative magnetic resonance imaging hippocampal
2011;93(2-3):96-106. abnormalities in patients undergoing temporal lobectomy for
medically refractory seizures. Epilepsia 1994;35:520-4.
Dunlea 2010 {published data only}
Dunlea O, Doherty CP, Farrell M, Fitzsimons M, O'Brien D, Gelinas 2011 {published data only}
Murphy K, et al. The Irish epilepsy surgery experience: long-term Gelinas JN, Battison AW, Smith S, Connolly MB, Steinbok P.
follow-up. Seizure 2010;19(4):247-52. Electrocorticography and seizure outcomes in children with
lesional epilepsy. Childs Nervous System 2011;27(3):381-90.
Elsharkawy 2008a {published data only}
Elsharkawy AE, Pannek H, Schulz R, Hoppe M, Pahs G, Georgakoulias 2008 {published data only}
Gyimesi C, et al. Outcome of extratemporal epilepsy surgery Georgakoulias NV, Mitsos AP, Konstantinou EA, Nicholson C,
experience of a single center. Neurosurgery 2008;63(3):516-25. Jenkins A. Trans-Sylvian selective amygdalohippocampectomy
for medically intractable temporal lobe epilepsy: a single-centre
Elsharkawy 2009a {published data only} experience. British Journal of Neurosurgery 2008;22(4):535-41.
Elsharkawy AE, Alabbasi AH, Pannek H, Oppel F, Schulz R,
Hoppe M, et al. Long-term outcome after temporal lobe Gilliam 1997a {published data only}
epilepsy surgery in 434 consecutive adult patients. Journal of Gilliam F, Bowling S, Bilir E, Thomas J, Faught E, Morawetz R, et
Neurosurgery 2009;110(6):1135-46. al. Association of combined MRI, interictal EEG, and ictal EEG
results with outcome and pathology after temporal lobectomy.
Elsharkawy 2011a {published data only} Epilepsia 1997;38:1315-20.
Elsharkawy AE, Pannek H, Woermann FG, Gyimesi C,
Hartmann S, Aengenendt J, et al. Apical temporal lobe Gilliam 1997b {published data only}
resection; "tailored" hippocampus-sparing resection Gilliam F, Wyllie E, Kashden J, Faught E, Kotagal P, Bebin M, et
based on presurgical evaluation data. Acta Neurochirurgica al. Epilepsy surgery outcome: comprehensive assessment in
2011;153(2):231-8. children. Neurology 1997;48:1368-74.
Engman 2004 {published data only} Goldstein 1996 {published data only}
Engman E, Andersson-Roswall L, Svensson E, Malmgren K. Goldstein R, Harvey AS, Duchowny M, Jayakar P, Altman N,
Non-parametric evaluation of memory changes at group Resnick T, et al. Preoperative clinical, EEG, and imaging
and individual level following temporal lobe resection for findings do not predict seizure. Journal of Child Neurology
pharmaco-resistant partial epilepsy. Journal of Clinical & 1996;11:445-50.
Experimental Neuropsychology 2004;26(7):943-54.
Greiner 2011 {published data only}
Erba 1992 {published data only} Greiner H, Park Y, Holland K, Horn P, Byars A, Mangano F, et al.
Erba G, Winston KR, Adler JR, Welch K, Ziegler R, Hornig GW. Scalp EEG does not predict hemispherectomy outcome. Seizure
Temporal lobectomy for complex partial seizures that began in 2011;20:758-63.
childhood. Surgical Neurology 1992;38:424-32.
Grivas 2006 {published data only}
Erickson 2005 {published data only} Grivas A, Schramm J, Kral T, von Lehe M, Helmstaedter C,
Erickson JC, Ellenbogen RG, Khajevi K, Mulligan L, Ford GC, Elger CE, et al. Surgical treatment for refractory temporal
Jabbari B. Temporal lobectomy for refractory epilepsy in the lobe epilepsy in the elderly: seizure outcome and
U.S. military. Military Medicine 2005;170(3):201-5. neuropsychological sequels compared with a younger cohort.
Epilepsia 2006;47(8):1364-72.
Fauser 2004 {published data only}
Fauser S, Bast T, Altenmuller DM, Schulte-Monting J, Strobl K, Gyimesi 2007 {published data only}
Steinhoff BJ, et al. Factors influencing surgical outcome in Gyimesi C, Fogarasi A, Kovacs N, Toth V, Magalova V, Schulz R,
patients with focal cortical dysplasia. Journal of Neurology, et al. Patients' ability to react before complex partial seizures.
Neurosurgery & Psychiatry 2008;79(1):103-5. Epilepsy & Behavior 2007;10(1):183-6.
Fujiwara 2012 {published data only} Hader 2004 {published data only}
Fujiwara H, Greiner HM, Lee KH, Holland-Bouley KD, Seo JH, Hader WJ, Mackay M, Otsubo H, Chitoku S, Weiss S, Becker L,
Arthur T, et al. Resection of ictal high-frequency oscillations et al. Cortical dysplastic lesions in children with intractable
leads to favorable surgical outcome in pediatric epilepsy. epilepsy: role of complete resection. Journal of Neurosurgery
Epilepsia 2012;53(9):1607-17. 2004;100(2 Suppl Pediatrics):110-7.
Informed decisions.

Hajek 2009 {published data only} frequency and postoperative seizures. Epilepsy Research
Hajek M, Krsek P, Dezortova M, Marusic P, Zamecnik J, Kyncl M, 2003;57(2-3):153-8.
et al. 1H MR spectroscopy in histopathological subgroups
Jaramillo-Betancur 2009 {published data only}
of mesial temporal lobe epilepsy. European Radiology
2009;19(2):400-8. Jaramillo-Betancur H, Jimenez ME, Massaro-Ceballos M, Cortes-
Silva E, Restrepo-Marin DM, Mora-Lopez O, et al. Risk factors
Hallbook 2010 {published data only} for seizure recurrence and short term outcome after epilepsy
Hallbook T, Ruggieri P, Adina C, Lachhwani DK, Gupta A, surgery for mesial temporal sclerosis [Factores de riesgo para
Kotagal P, et al. Contralateral MRI abnormalities in candidates recurrencia de convulsiones y prognostico a corto plazo en
for hemispherectomy for refractory epilepsy. Epilepsia cirugia de epilepsia para esclerosis mesial temporal]. Revista de
2010;51(4):556-63. Neurologia 2009;49:170-80.
Hamiwka 2005 {published data only} Jayakar 2008 {published data only}
Hamiwka L, Jayakar P, Resnick T, Morrison G, Ragheb J, Dean P, Jayakar P, Dunoyer C, Dean P, Ragheb J, Resnick T, Morrison G,
et al. Surgery for epilepsy due to cortical malformations: ten- et al. Epilepsy surgery in patients with normal or nonfocal
year follow-up. Epilepsia 2005;46(4):556-60. MRI scans: integrative strategies offer long-term seizure relief.
Epilepsia 2008;49(5):758-64.
Hartley 2002 {published data only}
Jayalakshmi 2011 {published data only}
Hartley LM, Gordon I, Harkness W, Harding B, Neville BG,
Cross JH. Correlation of SPECT with pathology and Jayalakshmi S, Panigrahi M, Kulkarni DK, Uppin M,
seizure outcome in children undergoing epilepsy surgery. Somayajula S, Challa S. Outcome of epilepsy surgery in children
Developmental Medicine & Child Neurology 2002;44(10):676-80. after evaluation with non-invasive protocol. Neurology India
2011;59(1):30-6.
Hartzfield 2008 {published data only}
Jeha 2006 {published data only}
Hartzfeld P, Elisevich K, Pace M, Smith B, Gutierrez JA.
Characteristics and surgical outcomes for medial temporal Jeha LE, Najm IM, Bingaman WE, Khandwala F, Widdess-
post-traumatic epilepsy. British Journal of Neurosurgery Walsh P, Morris HH, et al. Predictors of outcome after temporal
2008;22(2):224-30. lobectomy for the treatment of intractable epilepsy. Neurology
2006;66(12):1938-40.
Hemb 2010 {published data only}
Jehi 2012 {published data only}
Hemb M, Velasco TR, Parnes MS, Wu JY, Lerner JT,
Matsumoto JH, et al. Improved outcomes in pediatric Jehi L, Irwin A, Kayyali H, Vadera S, Bingaman W, Najm I.
epilepsy surgery: the UCLA experience, 1986-2008. Neurology Levetiracetam may favorably affect seizure outcome after
2010;74(22):1768-75. temporal lobectomy. Epilepsia 2012;53(6):979-86.
Holmes 1997 {published data only} Jennum 1993 {published data only}
Holmes MD, Dodrill CB, Ojemann GA, Wilensky AJ, Ojemann LM. Jennum P, Dhuna A, Davies K, Fiol M, Maxwell R. Outcome
Outcome following surgery in patients with bitemporal of resective surgery for intractable partial epilepsy guided
interictal epileptiform patterns. Neurology 1997;48:1037-40. by subdural electrode arrays. Acta Neurologica Scandinavica
1993;87:434-7.
Holmes 2000 {published data only}
Jeong 1999 {published data only}
Holmes MD, Kutsy RL, Ojemann GA, Wilensky AJ, Ojemann LM.
Interictal, unifocal spikes in refractory extratemporal epilepsy Jeong SW, Lee SK, Kim KK, Kim H, Kim JV, Chung CK. Prognostic
predict ictal origin and postsurgical outcome. Clinical factors in anterior temporal lobe resections for mesial temporal
Neurophysiology 2000;111:1802-8. lobe epilepsy: multivariate analysis. Epilepsia 1999;40:1735-9.
Jack 1992 {published data only} Kan 2008 {published data only}
Jack Jr. CR, Sharbrough FW, Cascino GD, Hirschorn KA, Kan P, Van Orman C, Kestle JR. Outcomes after surgery for focal
O’Brien PC, Marsh WR. Magnetic resonance image-based epilepsy in children. Child's Nervous System 2008;24(5):587-91.
hippocampal volumetry: correlation with outcome after
Kang 2009 {published data only}
temporal lobectomy. Annals of Neurology 1992;31:138-46.
Kang C, Cascino GD. The effect of preoperative body mass index
Janszky 2003a {published data only} on outcome after temporal lobe epilepsy surgery. Epilepsy
Janszky J, Schulz R, Ebner A. Clinical features and surgical Research 2009;87(2-3):272-6.
outcome of medial temporal lobe epilepsy with a history of
Kanner 2009 {published data only}
complex febrile convulsions. Epilepsy Research 2003;55(1-2):1-8.
Kanner AM, Byrne R, Chicharro A, Wuu J, Frey M. A lifetime
Janszky 2003b {published data only} psychiatric history predicts a worse seizure outcome following
Janszky J, Schulz R, Hoppe M, Ebner A. Spikes on the temporal lobectomy. Neurology 2009;72(9):793-9.
postoperative EEG are related to preoperative spike

Informed decisions.

Kilpatrick 1997 {published data only} Lee 2006 {published data only}
Kilpatrick C, Cook M, Kaye A, Murphy M, Matkovic Z. Non- Lee SA, Yim SB, Lim YM, Kang JK, Lee JK. Factors predicting
invasive investigations successfully select patients for temporal seizure outcome of anterior temporal lobectomy for patients
lobe surgery. Journal of Neurology Neurosurgery and Psychiatry with mesial temporal sclerosis. Seizure 2006;15(6):397-404.
1997;63:327-33.
Lee 2008 {published data only}
Kim 2009 {published data only} Lee JJ, Lee SK, Lee SY, Park KI, Kim DW, Lee DS, et al. Frontal
Kim DW, Lee SK, Chu K, Park KI, Lee SY, Lee CH, et al. Predictors lobe epilepsy: clinical characteristics, surgical outcomes and
of surgical outcome and pathologic considerations in focal diagnostic modalities. Seizure 2008;17(6):514-23.
cortical dysplasia. Neurology 2009;72(3):211-6.
Lee 2010a {published data only}
Kim 2010a {published data only} Lee YJ, Kang HC, Bae SJ, Kim HD, Kim JT, Lee BI, et al.
Kim DW, Kim HK, Lee SK, Chu K, Chung CK. Extent of neocortical Comparison of temporal lobectomies of children and adults
resection and surgical outcome of epilepsy: intracranial EEG with intractable temporal lobe epilepsy. Child's Nervous System
analysis. Epilepsia 2010;51(6):1010-7. 2010;26(2):177-83.
Kim 2010b {published data only} Lee 2011 {published data only}
Kim DW, Lee SK, Nam H, Chu K, Chung CK, Lee SY, et al. Lee JY, Phi JH, Wang KC, Cho BK, Kim SK. Transsylvian-
Epilepsy with dual pathology: surgical treatment of cortical transcisternal selective lesionectomy for pediatric lesional
dysplasia accompanied by hippocampal sclerosis. Epilepsia mesial temporal lobe epilepsy. Neurosurgery 2011;68(3):582-7.
2010;51(8):1429-35.
Lei 2008 {published data only}
Kloss 2002 {published data only} Lei T, Shu K, Chen X, Li L. Surgical treatment of epilepsy with
Kloss S, Pieper T, Pannek H, Holthausen H, Tuxhorn I. Epilepsy chronic cerebral granuloma caused by Schistosoma japonicum.
surgery in children with focal cortical dysplasia (FCD): results of Epilepsia 2008;49(1):73-9.
long-term seizure outcome. Neuropediatrics 2002;33(1):21-6.
Li 1997 {published data only}
Knowlton 2008 {published data only} Li LM, Cendes F, Watson C, Andermann F, Fish DR, Dubeau F,
Knowlton RC, Elgavish RA, Bartolucci A, Ojha B, Limdi N, et al. Surgical treatment of patients with single and dual
Blount J, et al. Functional imaging: II. Prediction of epilepsy pathology: relevance of lesion and of hippocampal atrophy to
surgery outcome. Annals of Neurology 2008;64(1):35-41. seizure outcome. Neurology 1997;48:437-44.
Kral 2007 {published data only} Li 1999 {published data only}

Kral T, von Lehe M, Podlogar M, Clusmann H, Sussmann P, Li LM, Cendes F, Andermann F, Watson C, Fish DR, Cook MJ,
Kurthen M, et al. Focal cortical dysplasia: long term seizure et al. Surgical outcome in patients with epilepsy and dual
outcome after surgical treatment. Journal of Neurology, pathology. Brain 1999;122:799-805.
Neurosurgery & Psychiatry 2007;78(8):853-6.
Liang 2010 {published data only}
Krsek 2013 {published data only} Liang S, Li A, Zhao M, Jiang H, Meng X, Sun Y. Anterior temporal
Krsek P, Kudr M, Jahodova A, Komarek V, Maton B, Malone S, lobectomy combined with anterior corpus callosotomy in
et al. Localizing value of ictal SPECT is comparable to MRI patients with temporal lobe epilepsy and mental retardation.
and EEG in children with focal cortical dysplasia. Epilepsia Seizure 2010;19(6):330-4.
2013;54(2):351-8.
Liang 2012 {published data only}
Kuzniecky 1993 {published data only} Liang S, Wang S, Zhang J, Ding C, Zhang Z, Fu X, et al. Long-
Kuzniecky R, Burgard S, Faught E, Morawets R, Bartolucci A. term outcomes of epilepsy surgery in school-aged children with
Predictive value of magnetic resonance imaging in temporal partial epilepsy. Pediatric Neurology 2012;47(4):284-90.
lobe epilepsy surgery. Archives of Neurology 1993;50:65-9.
Liava 2012 {published data only}
Kwan 2010 {published data only} Liava A, Francione S, Tassi L, Lo Russo G, Cossu M, Mai R, et al.
Kwan A, Ng WH, Otsubo H, Ochi A, Snead OC 3rd, Tamber MS, Individually tailored extratemporal epilepsy surgery in children:
et al. Hemispherectomy for the control of intractable epilepsy anatomo-electro-clinical features and outcome predictors in a
in childhood: comparison of 2 surgical techniques in a single population of 53. Epilepsy & Behavior 2012;25(1):68-80.
institution. Neurosurgery 2010;67(2):429-36.
Lopez-Gonzalez 2012 {published data only}
Lackmayer 2013 {published data only} Lopez-Gonzalez M, Gonzalez- Martinez J, Jehi L, Kotagal P,
Lackmayer K, Lehner-Baumgartner E, Pirker S, Czech T, Warbel A, Bingaman W. Epilepsy surgery of the temporal lobe
Baumgartner C. Preoperative depressive symptoms are not in pediatric population: a retrospective analysis. Neurosurgery
predictors of postoperative seizure control in patients with 2012;70:684-92.
mesial temporal lobe epilepsy and hippocampal sclerosis.
Epilepsy & Behavior 2013;26(1):81-6.
Informed decisions.

Lorenzo 1995 {published data only} O'Brien 2000 {published data only}
Lorenzo NY, Parisi JE, Cascino GD, Jack Jr. CR, Marsh WR, O’Brien TJ, So EL, Mullas BP, Cascino GD, Hauser MF,
Hirschorn KA. Intractable frontal lobe epilepsy: pathological and Brinkmann BH, et al. Subtraction peri-ictal SPECT is predictive
MRI features. Epilepsy Research 1995;20:171-8. of extratemporal epilepsy surgery outcome. Neurology
2000;55:1668-77.
Madhavan 2007 {published data only}
Madhavan D, Schaffer S, Yankovsky A, Arzimanoglou A, Oertel 2005 {published data only}
Renaldo F, Zaroff CM, et al. Surgical outcome in tuberous Oertel J, Gaab MR, Runge U, Schroeder HW, Piek J. Waterjet
sclerosis complex: a multicenter survey. Epilepsia dissection versus ultrasonic aspiration in epilepsy surgery.
2007;48(8):1625-8. Neurosurgery 2005;56(1 Suppl):142-6; discussion 142-6.
Mani 2006 {published data only} Paglioli 2006 {published data only}
Mani J, Gupta A, Mascha E, Lachhwani D, Prakash K, Paglioli E, Palmini A, Portuguez M, Paglioli E, Azambuja N,
Bingaman W, et al. Postoperative seizures after extratemporal da Costa JC, et al. Seizure and memory outcome following
resections and hemispherectomy in pediatric epilepsy. temporal lobe surgery: selective compared with nonselective
Neurology 2006;66(7):1038-43. approaches for hippocampal sclerosis. Journal of Neurosurgery
2006;104(1):70-8.
Mathern 1999 {published data only}
Mathern GW, Giza CC, Yudovin S, Vinters HV, Peacock WJ, Paolicchi 2000 {published data only}
Shewmon DA, et al. Postoperative seizure control and Paolicchi JM, Jayakar P, Dean P, Yaylali I, Morrison G, Prats A,
antiepileptic drug use in pediatric epilepsy surgery patients: the et al. Predictors of outcome in pediatric epilepsy surgery.
UCLA experience. Epilepsia 1999;40(12):1740-9. Neurology 2000;54:642-7.
McIntosh 2012 {published data only} Park 2002 {published data only}
McIntosh A, Averill C, Kalnins RL, Mitchell LA, Fabinyi G, Park K, Buchhalter J, McClelland R, Raffel C. Frequency
Jackson G, et al. Long-term seizure outcome and risk factors and significance of acute postoperative seizures following
for recurrence after extratemporal epilepsy surgery. Epilepsia epilepsy surgery in children and adolescents. Epilepsia
20102;53(6):970-8. 2002;43(8):874-81.
Mihara 2004 {published data only} Park 2006 {published data only}
Mihara T, Matsuda K, Tottori T, Otsubo T, Baba K, Nishibayashi H, Park CK, Kim SK, Wang KC, Hwang YS, Kim KJ, Chae JH, et
et al. Long-term seizure outcome following resective surgery al. Surgical outcome and prognostic factors of pediatric
at National Epilepsy Center in Shizuoka, Japan. Psychiatry & epilepsy caused by cortical dysplasia. Child's Nervous System
Clinical Neurosciences 2004;58(3):S22-5. 2006;22(6):586-92.
Miserocchi 2013 {published data only} Perego 2009 {published data only}
Miserocchi A, Cascardo B, Piroddi C, Fuschillo D, Cardinale F, Setoain Perego X, Carreno Delgado M, Rumia Arboix J, Donaire
Nobili L, et al. Surgery for temporal lobe epilepsy in children: Pedraza A, Bargallo Alabart N, Seres Roig E, et al. Peri-ictal
relevance of presurgical evaluation and analysis of outcome. SPECT in temporal lobe epilepsy: post-surgical evaluation.
Journal of Neurosurgery. Pediatrics 2013;11(3):256-67. Revista Espanola De Medicina Nuclear 2009;28(2):56-62.
Morino 2009 {published data only} Perry 2010 {published data only}
Morino M, Ichinose T, Uda T, Kondo K, Ohfuji S, Ohata K. Perry MS, Dunoyer C, Dean P, Bhatia S, Bavariya A, Ragheb J, et
Memory outcome following transsylvian selective al. Predictors of seizure freedom after incomplete resection in
amygdalohippocampectomy in 62 patients with hippocampal children. Neurology 2010;75(16):1448-53.
sclerosis. Journal of Neurosurgery 2009;110(6):1164-9.
Phi 2009 {published data only}
Morris 1998 {published data only} Phi JH, Kim SK, Cho BK, Lee SY, Park SY, Park SJ, et al. Long-term
Morris HH, Matkovic Z, Estes ML, Prayson RA, Comair YG, surgical outcomes of temporal lobe epilepsy associated with
Turnbull J, et al. Ganglioglioma and intractable epilepsy: clinical low-grade brain tumors. Cancer 2009;115(24):5771-9.
and neurophysiologic features and predictors of outcome after
surgery. Epilepsia 1998;39(3):303-13. Phi 2010 {published data only}
Phi JH, Cho BK, Wang KC, Lee JY, Hwang YS, Kim KJ, et al.
O'Brien 1996 {published data only} Longitudinal analyses of the surgical outcomes of pediatric
O’Brien TJ, Kilpatrick C, Murrie V, Vogrin S, Morris K, epilepsy patients with focal cortical dysplasia. Journal of
Cook MJ. Temporal lobe epilepsy caused by mesial temporal Neurosurgery: Pediatrics 2010;6:49-56.
sclerosis and temporal neocortical lesions. A clinical and
electroencephalographic study of 46 pathologically proven Pinheiro-Martins 2012 {published data only}
cases. Brain 1996;119:2133-41. Pinheiro-Martinsa A, Bianchinc M, Velascoa T, Terraa V, Araújo
Jr. D, Wichert-Anaa L, et al. Independent predictors and a

Informed decisions.

prognostic model for surgical outcome in refractory frontal lobe Salanova 1994 {published data only}
epilepsy. Epilepsy Research 2012;99:55-63. Salanova V, Markand ON, Worth R. Clinical characteristics
and predictive factors in 98 patients with complex partial
Prevedello 2000 {published data only}
seizures treated with temporal resection. Archives of Neurology
Prevedello DMS, Sandmann MC, Ebner A. Prognostic factors 1994;51:1008-13.
in mesial temporal lobe epilepsy surgery. Arquivos de Neuro-
Psiquiatria 2000;58(2-A):207-13. Sarkis 2012 {published data only}
Sarkis R, Jehi L, Najm I, Kotagal P, Bingaman W. Seizure
Raabe 2012 {published data only}
outcomes following multilobar epilepsy surgery. Epilepsia
Raabe A, Schmitz A, Pernhorst K, Grote A, von der Brelie C, 2010;53(1):44-50.
Urbach H, et al. Cliniconeuropathologic correlations show
astroglial albumin storage as a common factor in epileptogenic Schramm 2011 {published data only}
vascular lesions. Epilepsia 2012;53(3):539-48. Schramm J, Lehmann TN, Zentner J, Mueller CA, Scorzin J,
Fimmers R, et al. Randomized controlled trial of 2.5-cm
Radhakrishnan 1998 {published data only}
versus 3.5-cm mesial temporal resection in temporal lobe
Radhakrishnan K, So EL, Silbert PL, Jack Jr. CR, Cascino GD, epilepsy--Part 1: intent-to-treat analysis. Acta Neurochirurgica
Sharbrough FW, et al. Predictors of outcome of anterior 2011;153(2):209-19.
temporal lobectomy for intractable epilepsy: a multivariate
study. Neurology 1998;51:465-71. Schramm J, Lehmann TN, Zentner J, Mueller CA, Scorzin J,
Fimmers R, et al. Randomized controlled trial of 2.5-cm
Rausch 2003 {published data only} versus 3.5-cm mesial temporal resection in temporal lobe
Rausch R, Kraemer S, Pietras CJ, Le M, Vickrey BG, Passaro EA. epilepsy--Part 2: volumetric analysis. Acta Neurochirurgica
Early and late cognitive changes following temporal lobe 2011;153(2):221-8.
surgery for epilepsy. Neurology 2003;60(6):951-9.
Seymour 2012 {published data only}
Remi 2011 {published data only} Seymour N, Granbichler CA, Polkey CE, Nashef L. Mortality after
Remi J, Vollmar C, de Marinis A, Heinlin J, Peraud A, Noachtar S. temporal lobe epilepsy surgery. Epilepsia 2012;53(2):267-71.
Congruence and discrepancy of interictal and ictal EEG with MRI
lesions in focal epilepsies. Neurology 2011;77(14):1383-90. Sinclair 2003 {published data only}
Sinclair DB, Aronyk KE, Snyder TJ, Wheatley BM, McKean JD,
Roberti 2007 {published data only} Bhargava R, et al. Pediatric epilepsy surgery at the University of
Roberti F, Potolicchio SJ, Caputy AJ. Tailored anteromedial Alberta: 1988-2000. Pediatric Neurology 2003;29(4):302-11.
lobectomy in the treatment of refractory epilepsy of the
temporal lobe: long term surgical outcome and predictive Sindou 2006 {published data only}
factors. Clinical Neurology & Neurosurgery 2007;109(2):158-65. Sindou M, Guenot M, Isnard J, Ryvlin P, Fischer C, Mauguiere F.
Temporo-mesial epilepsy surgery: outcome and complications
Rossi 1994 {published data only} in 100 consecutive adult patients. Acta Neurochirurgica
Rossi GF, Colicchio G, Scerrati M. Resection surgery for partial 2006;148(1):39-45.
epilepsy. Relation of surgical outcome with some aspects
of the epileptogenic process and surgical approach. Acta Sola 2005 {published data only}
Neurochirurgica 1994;130:101-10. Sola RG, Hernando-Requejo V, Pastor J, Garcia-Navarrete E,
DeFelipe J, Alijarde MT, et al. Pharmacoresistant temporal-
Russo 2003 {published data only} lobe epilepsy. Exploration with foramen ovale electrodes
Russo GL, Tassi L, Cossu M, Cardinale F, Mai R, Castana L, and surgical outcomes [Epilepsia farmacorresistemente del
et al. Focal cortical resection in malformations of cortical lobulo temporal. Exploracion con electrodos del foramen oval y
development. Epileptic Disorders 2003;5(Suppl 2):S115-23. resultados quirurgicos]. Revista de Neurologia 2005;41:4-16.
Sagher 2012 {published data only} Spencer 2005 {published data only}
Sagher O, Thawani J, Etame A, Gomez-Hassan D. Seizure Spencer SS, Berg AT, Vickrey BG, Sperling MR, Bazil CW,
outcomes and mesial resection volumes following selective Shinnar S, et al. Initial outcomes in the Multicenter Study of
amygdalohippocampectomy and temporal lobectomy. Epilepsy Surgery. Neurology 2003;61(12):1680-5.
Neurosurgery Focus 2012;32(3; E8):1-8.
Spencer SS, Berg AT, Vickrey BG, Sperling MR, Bazil CW,
Sakamoto 2009 {published data only} Shinnar S, et al. Predicting long-term seizure outcome after
Sakamoto S, Takami T, Tsuyuguchi N, Morino M, Ohata K, resective epilepsy surgery: the multicenter study. Neurology
Inoue Y, et al. Prediction of seizure outcome following epilepsy 2005;65(6):912-8.
surgery: asymmetry of thalamic glucose metabolism and
Sperling 1992 {published data only}
cerebral neural activity in temporal lobe epilepsy. Seizure
2009;18(1):1-6. Sperling MR, O’Connor MJ, Saykin AJ, Phillips CA, Morrell MJ,
Bridgman PA, et al. A noninvasive protocol for anterior temporal
lobectomy. Neurology 1992;42:416-22.

Informed decisions.

Stavrou 2008 {published data only} Tripathi 2008 {published data only}
Stavrou I, Baumgartner C, Frischer JM, Trattnig S, Knosp E. Tripathi M, Singh MS, Padma MV, Gaikwad S, Bal CS, Tripathi M,
Long-term seizure control after resection of supratentorial et al. Surgical outcome of cortical dysplasias presenting with
cavernomas: a retrospective single-center study in 53 patients. chronic intractable epilepsy: a 10-year experience. Neurology
Neurosurgery 2008;63(5):888-96. India 2008;56(2):138-43.
Suppiah 2009 {published data only} Trottier 2008 {published data only}
Suppiah R, Mee E, Walker EB, Roberts L, Finucane G, Bergin PS. Trottier S, Landre E, Biraben A, Chassoux F, Pasnicu A,
Temporal lobe resection for refractory temporal lobe Scarabin JM, et al. On the best strategies on the best results
epilepsy at Auckland Hospital. New Zealand Medical Journal for surgery of frontal epilepsy [Quelles startegies pour quels
2009;12(1305):47-56. resultats dans la chirurgerie des epliepsies partielles frontales?].
Neuro-Chirurgie 2008;54(3):388-98.
Swartz 1992 {published data only}
Swartz BE, Tomiyasu U, Delgado-Escueta AV, Mandelkern M, Urbach 2007 {published data only}
Khonsari A. Neuroimaging in temporal lobe epilepsy: test Urbach H, Binder D, von Lehe M, Podlogar M, Bien CG, Becker A,
sensitivity and relationships to pathology and postoperative et al. Correlation of MRI and histopathology in epileptogenic
outcome. Epilepsia 1992;33:624-34. parietal and occipital lobe lesions. Seizure 2007;16(7):608-14.
Tanriverdi 2010 {published data only} Ure 2009 {published data only}
Tanriverdi T, Dudley RW, Hasan A, Al Jishi A, Al Hinai Q, Ure JA, Olivier A, Quesney LF, Bravo M, Perassolo M.
Poulin N, et al. Memory outcome after temporal lobe epilepsy Temporolimbic activation by intracranial electrical stimulation.
surgery: corticoamygdalohippocampectomy versus selective Canadian Journal of Neurological Sciences 2009;36(5):593-8.
amygdalohippocampectomy. Journal of Neurosurgery
2010;113(6):1164-75. Velasco 2011 {published data only}
Velasco TR, Wichert-Ana L, Mathern GW, Araújo D, Walz R,
Tatum 2008 {published data only} Bianchin MM, et al. Utility of ictal single photon emission
Tatum WO 4th, Benbadis SR, Hussain A, Al-Saadi S, Kaminski B, computed tomography in mesial temporal lobe epilepsy
Heriaud LS, et al. Ictal EEG remains the prominent predictor with hippocampal atrophy: a randomized trial. Neurosurgery
of seizure-free outcome after temporal lobectomy in epileptic 2011;68:431-6.
patients with normal brain MRI. Seizure 2008;17(7):631-6.
Walz 2003 {published data only}
Terra-Bustamante 2005a {published data only} Walz R, Castro RM, Velasco TR, Alexandre V Jr, Lopes MH,
Terra-Bustamante VC, Fernandes RM, Inuzuka LM, Velasco TR, Leite JP, et al. Surgical outcome in mesial temporal sclerosis
Alexandre V Jr, Wichert-Ana L, et al. Surgically amenable correlates with prion protein gene variant. Neurology
epilepsies in children and adolescents: clinical, imaging, 2003;61(9):1204-10.
electrophysiological, and post-surgical outcome data. Child's
Nervous System 2005;21(7):546-51. Weinand 1992 {published data only}
Weinand ME, Wyler AR, Richey ET, Phillips BB, Somes GW. Long-
Terra-Bustamante 2005b {published data only} term ictal monitoring with subdural strip electrodes: prognostic
Terra-Bustamante VC, Inuzuca LM, Fernandes RM, Funayama S, factors for selecting temporal lobectomy candidates. Journal of
Escorsi-Rosset S, Wichert-Ana L, et al. Temporal lobe epilepsy Neurosurgery 1992;77:20-8.
surgery in children and adolescents: clinical characteristics and
post-surgical outcome. Seizure 2005;14(4):274-81. Wellmer 2012 {published data only}
Wellmer J, von der Groeben F, Klarmann U, Weber C, Elger C,
Tezer 2008 {published data only} Urbach H, et al. Risks and benefits of invasive epilepsy surgery
Tezer FI, Akalan N, Oguz KK, Karabulut E, Dericioglu N, Ciger A, workup with implanted subdural and depth electrodes.
et al. Predictive factors for postoperative outcome in temporal Epilepsia 2012;53(8):1322-32.
lobe epilepsy according to two different classifications. Seizure
2008;17(6):549-60. Widdess-Walsh 2007 {published data only}
Widdess-Walsh P, Jeha L, Nair D, Kotagal P, Bingaman W,
Theodore 2012 {published data only} Najm I. Subdural electrode analysis in focal cortical dysplasia:
Theodore WH, Martinez AR, Khan OI, Liew CJ, Auh S, Dustin IM, predictors of surgical outcome. Neurology 2007;69(7):660-7.
et al. PET of serotonin 1A receptors and cerebral glucose
metabolism for temporal lobectomy. Journal of Nuclear Wiebe 2001 {published data only}
Medicine 2012;53(9):1375-82. Wiebe S, Blume WT, Girvin JP, Eliasziw M. A randomized,
controlled trial of surgery for temporal-lobe epilepsy. New
Tigaran 2003 {published data only} England Journal of Medicine 2001;345(5):311-8.
Tigaran S, Cascino GD, McClelland RL, So EL, Richard Marsh W.
Acute postoperative seizures after frontal lobe cortical resection
for intractable partial epilepsy. Epilepsia 2003;44(6):831-5.

Informed decisions.

Wieshmann 2008 {published data only} patients. Journal of Neurology, Neurosurgery and Psychiatry
Wieshmann UC, Larkin D, Varma T, Eldridge P. Predictors of 1995;58:666-73.
outcome after temporal lobectomy for refractory temporal lobe
Zentner 1996 {published data only}
epilepsy. Acta Neurologica Scandinavica 2008;118(5):306-12.
Zentner J, Hufnagel A, Ostertun B, Wolf HK, Behrens E,
Wray 2012 {published data only} Campos MG, et al. Surgical treatment of extratemporal epilepsy:
Wray CD, McDaniel SS, Saneto RP, Novotny EJ Jr, Ojemann JG. clinical, radiologic and histopathologic findings in 60 patients.
Is postresective intraoperative electrocorticography predictive Epilepsia 1996;37:1072-80.
of seizure outcomes in children?. Journal of Neurosurgery.

Pediatrics 2012;9(5):1546-51.
References to studies excluded from this review
Wyler 1995 {published data only} Acar 2008 {published data only}
Wyler AR, Hermann BP, Somes G. Extent of medial temporal Acar G, Acar F, Miller, J, Spencer DC, Burchiel KJ.
resection on outcome from anterior temporal lobectomy: a Seizure outcome following transcortical selective
randomized prospective study. Neurosurgery 1995;37:982-90. amygdalohippocampectomy in mesial temporal lobe epilepsy.
Stereotactic & Functional Neurosurgery 2008;86(5):314-9.
Wyllie 1998 {published data only}
Wyllie E, Comair YG, Kotagal P, Bulacio J, Bingaman W, Alemany-Rosales 2011 {published data only}
Ruggirei P. Seizure outcome after epilepsy surgery in children Alemany-Rosales B, Prieto-Montalvo J. Refractory temporal
and adolescents. Annals of Neurology 1998;44:740-8. epilepsy. An analysis of 33 cases submitted to surgery [Epilepsia
temporal farmacorresistante. Analisis de 33 casos intervenidos
Yang 2011 {published data only}
quirurgicamente]. Revista de Neurologia 2011;52(10):581-9.
Yang XL, Lu QC, Xu JW, Wang GS, Liu Q. Predictors of outcome
in the surgical treatment for epilepsy. Chinese Medical Journal Alpherts 2008 {published data only}
2011;124(24):4166-71. * Alpherts WC, Vermeulen J, van Rijen PC, Lopes da Silva FH,
van Veelen WM, on behalf of the Dutch Collaborative Epilepsy
Yeon 2009 {published data only}
Surgery Program. Verbal memory decline after temporal
Yeon JY, Kim JS, Choi SJ, Seo DW, Hong SB, Hong SC. epilepsy surgery?: A 6-year multiple assessments follow-up
Supratentorial cavernous angiomas presenting with seizures: study. Neurology 2006;67(4):626-31.
surgical outcomes in 60 consecutive patients. Seizure
2009;18(1):14-20. Andersson-Roswall 2010 {published data only}
Andersson-Roswall L, Engman E, Samuelsson H, Malmgren K.
Yu 2009 {published data only}
Cognitive outcome 10 years after temporal lobe surgery.
Yu T, Wang Y, Zhang G, Cai L, Du W, Li Y. Posterior cortex Neurology 2011;74:1977-85.
epilepsy: diagnostic considerations and surgical outcome.
Seizure 2009;18(4):288-92. Asadi-Pooya 2008 {published data only}
Asadi-Pooya AA, Nei M, Sharan AD, Mintzer S, Zangaladze A,
Yu 2012a {published data only}
Evans JG, et al. Antiepileptic drugs and relapse after epilepsy
Yu T, Zhang G, Kohrman MH, Wang Y, Cai L, Shu W, et al. A surgery. Epileptic Disorders 2008;10(3):193-8.
retrospective study comparing preoperative evaluations and
postoperative outcomes in paediatric and adult patients Bauer 2007 {published data only}
undergoing surgical resection for refractory epilepsy. Seizure Bauer R, Dobesberger J, Unterhofer C, Unterberger I, Walser G,
2012;21(6):444-9. Bauer G, et al. Outcome of adult patients with temporal lobe
tumours and medically refractory focal epilepsy. Journal of
Yu 2012b {published data only}
Neurosurgery 2011;7(2):179-88.
Yu T, Zhang G, Kohrman MH, Wang Y, Cai L, Shu W, et al. A
retrospective study comparing preoperative evaluations and Baxendale 2005 {published data only}
postoperative outcomes in paediatric and adult patients Baxendale S, Thompson P. Defining meaningful postoperative
undergoing surgical resection for refractory epilepsy. Seizure change in epilepsy surgery patients: measuring the
2012;21(6):444-9. unmeasurable?. Epilepsy & Behavior 2005;6(2):202-11.
Zangaladze 2008 {published data only} Bell 2010 {published data only}
Zangaladze A, Nei M, Liporace JD, Sperling MR. Characteristics Bell GS, Sinha S, Tisi J, Stephani C, Scott CA, Harkness WF, et
and clinical significance of subclinical seizures. Epilepsia al. Premature mortality in refractory partial epilepsy: does
2008;49(12):2016-21. surgical treatment make a difference?. Journal of Neurology,
Zentner 1995 {published data only}
Zentner J, Hufnagel A, Wolf HK, Ostertun B, Behrens E,
Campos MG, et al. Surgical treatment of temporal lobe epilepsy:
clinical, radiological and histopathological findings in 178

Informed decisions.

Binder 2009 {published data only} Choi 2004b {published data only}
Binder DK, Von Lehe M, Kral T, Bien CG, Urbach H, Schramm J, Choi YL, Park SH, Chung CK, Suh YL, Chi JG. Cortical laminar
et al. Surgical treatment of occipital lobe epilepsy. Journal of disorganization is correlated with severity of the cortical
Neurosurgery 2008;109(1):57-69. dysplasia and surgical outcome--experience with 150 cases.
Epilepsy Research 2004;62(2-3):171-8.
Boesebeck 2002 {published data only}
Boesebeck F, Schulz R, May T, Ebner A. Lateralizing semiology Cohen-Gadol 2003 {published data only}
predicts the seizure outcome after epilepsy surgery in the Cohen-Gadol AA, Wilhelmi BG, Collignon F, White JB, Britton JW,
posterior cortex. Brain 2002;125(Pt 10):2320-31. Cambier DM, et al. Long-term outcome of epilepsy surgery
among 399 patients with nonlesional seizure foci including
Boshuisen 2012 {published data only} mesial temporal lobe sclerosis. Journal of Neurosurgery
Boshuisen K, Arzimanoglou A, Cross JH, Uiterwaal CS, Polster T, 2006;104(4):513-24.
van Nieuwenhuizen O, et al. TimeToStop study group. Timing of
antiepileptic drug withdrawal and long-term seizure outcome Colonnelli 2012 {published data only}
after paediatric epilepsy surgery (TimeToStop): a retrospective Colonnelli MC, Cross JH, Davies S, D'Argenzio L, Scott RC,
observational study. Lancet Neurology 2012;11(9):784-91. Pickles A, et al. Psychopathology in children before and after
surgery for extratemporal lobe epilepsy. Developmental
Bourgeois 2007 {published data only} Medicine & Child Neurology 2012;54(6):521-6.
Bourgeois M, Crimmins DW, de Oliveira RS, Arzimanoglou A,
Garnett M, Roujeau T, et al. Surgical treatment of epilepsy in Coutin-Churchman 2012 {published data only}
Sturge-Weber syndrome in children. Journal of Neurosurgery Coutin-Churchman PE, Wu JY, Chen LL, Shattuck K, Dewar S,
2007;106(1 Suppl):20-8. Nuwer MR. Quantification and localization of EEG interictal
spike activity in patients with surgically removed epileptogenic
Buckingham 2010 {published data only} foci. Clinical Neurophysiology 2012;123(3):471-85.
Buckingham SE, Chervoneva I, Sharan A, Zangaladze A,
Mintzer S, Skidmore C, et al. Latency to first seizure after Cukiert 2009 {published data only}
temporal lobectomy predicts long-term outcome. Epilepsia Cukiert A, Burattini JA, Mariani PR, Cukiert CM, Argentoni-
2010;51(10):1987-93. Baldochi M, Baise-Zung C, et al. Outcome after extended
callosal section in patients with primary idiopathic generalized
Busch 2011 {published data only} epilepsy. Epilepsia 2009;50(6):1377-80.
Busch RM, Dulay MF, Kim KH, Chapin JS, Jehi L, Kalman CC, et
al. Pre-surgical mood predicts memory decline after anterior D'Angelo 2006 {published data only}
temporal lobe resection for epilepsy. Archives of Clinical D'Angelo VA, De Bonis C, Amoroso R, Cali A, D'Agruma L,
Neuropsychology 2011;26(8):739-45. Guarnieri V, et al. Supratentorial cerebral cavernous
malformations: clinical, surgical, and genetic involvement.
Caicoya 2007 {published data only} Neurosurgical Focus 2006;21(1):e9; 1-7.
Caicoya AG, Macarron J, Albisua J, Serratosa JM. Tailored
resections in occipital lobe epilepsy surgery guided by D'Argenzio 2011 {published data only}
monitoring with subdural electrodes: characteristics and D'Argenzio L, Colonnelli MC, Harrison S, Jacques TS,
outcome. Epilepsy Research 2007;77(1):1-10. Harkness W, Vargha-Khadem F, et al. Cognitive outcome
after extratemporal epilepsy surgery in childhood. Epilepsia
Carne 2004 {published data only} 2011;52(11):1966-72.
Carne RP, O'Brien TJ, Kilpatrick CJ, MacGregor LR, Hicks RJ,
Murphy MA, et al. MRI-negative PET-positive temporal lobe da Costa-Neves 2012 {published data only}
epilepsy: a distinct surgically remediable syndrome. Brain da Costa Neves RS, Jardim AP, Caboclo LO, Lancellotti C,
2004;127(10):2276-85. Marinho TF, Hamad AP, et al. Granule cell dispersion is not
a predictor of surgical outcome in temporal lobe epilepsy
Cascino 1996 {published data only} with mesial temporal sclerosis. Clinical Neuropathology
Cascino GD, Trenerry MR, So EL, Sharbrough FW, Shin C, 2013;32(1):24-30.
Lagerlund TD, et al. Routine EEG and temporal lobe epilepsy:
relation to long-term EEG monitoring, quantitative MRI, and Danielsson 2009 {published data only}
operative outcome. Epilepsia 1995;37:651-6. Danielsson S, Viggedal G, Steffenburg S, Rydenhag B, Gillberg C,
Olsson I. Psychopathology, psychosocial functioning, and IQ
Chang 2007 {published data only} before and after epilepsy surgery in children with drug-resistant
Chang EF, Wang DD, Barkovich AJ, Tihan T, Auguste KI, epilepsy. Epilepsy & Behavior 2009;14(2):330-7.
Sullivan JE, et al. Predictors of seizure freedom after surgery for
malformations of cortical development. Annals of Neurology Datta 2009 {published data only}
2011;70:151-62. Datta A, Sinclair DB, Wheatley M, Jurasek L, Snyder T, Quigley D,
et al. Selective amygdalohippocampectomy: surgical outcome
in children versus adults. Canadian Journal of Neurological
Sciences 2009;36(2):187-91.

Informed decisions.

Dulay 2006 {published data only} Griffin 2007 {published data only}
Dulay MF, York MK, Soety EM, Hamilton WJ, Mizrahi EM, Griffin S, Tranel D. Age of seizure onset, functional
Goldsmith IL, et al. Memory, emotional and vocational reorganization, and neuropsychological outcome in temporal
impairments before and after anterior temporal lobectomy for lobectomy. Journal of Clinical & Experimental Neuropsychology
complex partial seizures. Epilepsia 2006;47(11):1922-30. 2007;29(1):13-24.
Dulay 2009 {published data only} Grunert 2003 {published data only}
Dulay MF, Levin HS, York MK, Li X, Mizrahi EM, Goldsmith I, et Grunert P, Charalampaki K, Kassem M, Boecher-Schwarz H,
al. Changes in individual and group spatial and verbal learning Filippi R, Grunert P Jr. Frame-based and frameless stereotaxy in
characteristics after anterior temporal lobectomy. Epilepsia the localization of cavernous angiomas. Neurosurgical Review
2009;50(6):1385-95. 2003;26(1):53-61.
Elsharkawy 2008b {published data only} Haegelen 2013 {published data only}
Elsharkawy AE, Alabbasi AH, Pannek H, Schulz R, Hoppe M, Haegelen C, Perucca P, Châtillon CE, Andrade-Valença L,
Pahs G, et al. Outcome of frontal lobe epilepsy surgery in adults. Zelmann R, Jacobs J, et al. High-frequency oscillations, extent
Epilepsy Research 2008;81:97-106. of surgical resection, and surgical outcome in drug-resistant
focal epilepsy. Epilepsia 2013;54(5):848-57.
Elsharkawy 2009b {published data only}
Elsharkawy AE, El-Ghandour NM, Oppel F, Pannek H, Schulz R, Harvey 2008 {published data only}
Hoppe M, et al. Long-term outcome of lesional posterior cortical Harvey DJ, Naugle RI, Magleby J, Chapin JS, Najm IM,
epilepsy surgery in adults. Journal of Neurology, Neurosurgery & Bingaman W, et al. Relationship between presurgical memory
Psychiatry 2009;80(7):773-80. performance on the Wechsler Memory Scale-III and memory
change following temporal resection for treatment of
Elsharkawy 2011b {published data only} intractable epilepsy. Epilepsy & Behavior 2008;13(2):372-5.
Elsharkawy AE, Pietilä TA, Alabbasi AH, Pannek H, Ebner A. Long
term outcome in patients not initially seizure free after resective Hellwig 2012 {published data only}
epilepsy surgery. Seizure 2011;20(5):419-24. Hellwig S, Mamalis P, Feige B, Schulze-Bonhage A, van Elst LT.
Psychiatric comorbidity in patients with pharmacoresistant
Engel Jr 2012 {published data only} focal epilepsy and psychiatric outcome after epilepsy surgery.
Engel Jr J, McDermott M, Wiebe S, Langfitt J, Stern J, Dewar S, Epilepsy & Behavior 2012;23(2):272-9.
et al. Early surgical therapy for drug-resistant temporal lobe
epilepsy. JAMA 2012;307(9):922-30. Helmstaedter 2004 {published data only}
Helmstaedter C, Van Roost D, Clusmann H, Urbach H, Elger CE,
Fauser 2008 {published data only} Schramm J. Collateral brain damage, a potential source
Fauser S, Bast T, Altenmuller DM, Schulte-Monting J, Strobl K, of cognitive impairment after selective surgery for control
Steinhoff BJ, et al. Factors influencing surgical outcome in of mesial temporal lobe epilepsy. Journal of Neurology,
patients with focal cortical dysplasia. Journal of Neurology, Neurosurgery & Psychiatry 2004;75(2):323-6.
Helmstaedter 2011 {published data only}
Ferrari-Marinho 2012 {published data only} Helmstaedter C, Roeske S, Kaaden S, Elger C, Schramm J.
Ferrari-Marinho T, Caboclo LO, Marinho MM, Centeno RS, Hippocampal resection length and memory outcome in
Neves RS, Santana MT, et al. Auras in temporal lobe epilepsy selective epilepsy surgery. Journal of Neurology, Neurosurgery
with hippocampal sclerosis: relation to seizure focus and Psychiatry 2011;82:1375-81.
laterality and post surgical outcome. Epilepsy & Behavior
2012;24(1):120-5. Hervas-Navidad 2002 {published data only}
Hervas-Navidad R, Altuzarra-Corral A, Lucena-Martin JA,
Ferroli 2006 {published data only} Castaneda-Guerrero M, Vela-Yebra R, Sanchez-A lvarez JC.
Ferroli P, Casazza M, Marras C Mendola C, Franzini A, Broggi G. Defects in the visual field in resective surgery for temporal lobe
Cerebral cavernomas and seizures: a retrospective study on epilepsy [Defectos del campo visual en la cirurgia resectiva
163 patients who underwent pure lesionectomy. Neurological de la epilpesia del lobulo temporal]. Revista de Neurologia
Sciences 2006;26(6):390-4. 2002;34(11):1025-30.
Freitag 2005 {published data only} Hildebrandt 2005 {published data only}
Freitag H, Tuxhorn I. Cognitive function in preschool children Hildebrandt M, Schulz R, Hoppe M, May T, Ebner A.
after epilepsy surgery: rationale for early intervention. Epilepsia Postoperative routine EEG correlates with long-term seizure
2005;46(4):561-7. outcome after epilepsy surgery. Seizure 2005;14(7):446-51.
Ghacibeh 2009 {published data only} Hu 2012 {published data only}

Ghacibeh GA, Smith JD, Roper SN, Gilmore R, Eisenschenk S. Hu WH, Ge M, Zhang K, Meng FG, Zhang JG. Seizure outcome
Seizure recurrence following epilepsy surgery: is post-operative with surgical management of epileptogenic ganglioglioma: a
EEG helpful?. Seizure 2009;18(3):193-6. study of 55 patients. Acta Neurochirurgica 2012;154(5):855-61.

Informed decisions.

Jehi 2010 {published data only} examination of race, age, sex, and insurance status as
Jehi LE, Silveira DC, Bingaman W, Najm I. Temporal lobe factors predicting receipt of resective treatment. Journal of
epilepsy surgery failures: predictors of seizure recurrence, yield Neurosurgery 2007;107(6 Suppl):469-73.
of reevaluation, and outcome following reoperation. Journal of
McClelland 2011 {published data only}
Neurosurgery 2010;116(6):1186-94.
McClelland S 3rd, Guo H, Okuyemi KS. Population-based
Junna 2013 {published data only} analysis of morbidity and mortality following surgery for
Junna MR, Buechler R, Cohen-Gadol AA, Mandrekar J, intractable temporal lobe epilepsy in the United States. Archives
Christianson T, Marsh WR, et al. Prognostic importance of of Neurology 2011;68(6):725-9.
risk factors for temporal lobe epilepsy in patients undergoing
Mikati 2004 {published data only}
surgical treatment. Mayo Clinic Proceedings 2013;88(5):332-6.
Mikati MA, Comair Y, Ismail R, Faour R, Rahi AC. Effects of
Kuzniecky 1996 {published data only} epilepsy surgery on quality of life: a controlled study in a Middle
Kuzniecky RI, Burgard S, Bilir E, Morawetz R, Gilliam F, Faught E, Eastern population. Epilepsy & Behavior 2004;5(1):72-80.
et al. Qualitative MRI segmentation in mesial temporal sclerosis:
Mohammed 2012 {published data only}
clinical correlations. Epilepsia 1996;37:433-9.
Mohammed HS, Kaufman CB, Limbrick DD, Steger-May K,
Lach 2010 {published data only} Grubb RL Jr, Rothman SM, et al. Impact of epilepsy surgery on
Lach LM, Elliott I, Giecko T, Olds J, Snyder T, McCleary L, et seizure control and quality of life: a 26-year follow-up study.
al. Patient-reported outcome of pediatric epilepsy surgery: Epilepsia 2012;53(4):712-20.
social inclusion or exclusion as young adults?. Epilepsia
Moien-Afshari 2009 {published data only}
2010;51(10):2089-97.
Moien-Afshari F, Griebel R, Sadanand V, Vrbancic M, Hernandez-
Lachhwani 2003 {published data only} Ronquillo L, Lowry N, et al. Safety and yield of early cessation of
Lachhwani R, Luders H. In refractory temporal lobe epilepsy, AEDs in video-EEG telemetry and outcomes. Canadian Journal
consider surgery sooner. Cleveland Clinic Journal of Medicine of Neurological Sciences 2009;36(5):587-92.
2003;70(7):649-53.
Negishi 2011 {published data only}
Lee 2010 {published data only} Negishi M, Martuzzi R, Novotny EJ, Spencer DD, Constable RT.
Lee YJ, Kang HC, Lee JS, Kim SH, Kim DS, Shim KW, et al. Functional MRI connectivity as a predictor of the surgical
Resective pediatric epilepsy surgery in Lennox-Gastaut outcome of epilepsy. Epilepsia 2011;52(9):1733-40.
syndrome. Pediatrics 2010;125(1):e58-66.
Nikase 2007 {published data only}
Limbrick 2009 {published data only} Nakase H, Tamura K, Kim YJ, Hirabayashi H, Sakaki T, Hoshida T.
Limbrick DD, Narayan P, Powers AK, Ojemann JG, Park TS, Long-term follow-up outcome after surgical treatment for
Bertrand M, et al. Hemispherotomy: efficacy and analysis lesional temporal lobe epilepsy. Neurological Research
of seizure recurrence. Journal of Neurosurgery. Pediatrics 2007;29(6):588-93.
2009;4(4):323-32.
Oertel 2004 {published data only}
Lodenkemper 2007 {published data only} Oertel J, Gaab MR, Runge U, Schroeder HW, Wagner W, Piek J.
Loddenkemper T, Holland KD, Stanford LD, Kotagal P, Neuronavigation and complication rate in epilepsy surgery.
Bingaman W, Wyllie E. Developmental outcome after epilepsy Neurosurgical Review 2004;27(3):214-7.
surgery in infancy. Pediatrics 2007;119(5):930-5.
Ogiwara 2010 {published data only}
Lutz 2004 {published data only} Ogiwara H, Nordli DR, DiPatri AJ, Alden TD, Bowman RM,
Lutz MT, Clusmann H, Elger CE, Schramm J, Helmstaedter C. Tomita T. Pediatric epileptogenic gangliogliomas: seizure
Neuropsychological outcome after selective outcome and surgical results. Journal of Neurosurgery.
amygdalohippocampectomy with transsylvian versus Pediatrics 2010;5(3):271-6.
transcortical approach: a randomized prospective clinical
Park 2010 {published data only}
trial of surgery for temporal lobe epilepsy. Epilepsia
2004;45(7):809-16. Park KI, Lee SK, Chu K, Jung KH, Bae EK, Kim JS, et al.
Withdrawal of antiepileptic drugs after neocortical epilepsy
Malla 1998 {published data only} surgery. Annals of Neurology 2010;67(2):230-8.
Malla B, O’Brien TJ, Cascino GD, So EL, Radhakrishnan K,
Placantonakis 2010 {published data only}
Silbert P, et al. Acute postoperative seizures following anterior
temporal lobectomy for intractable partial epilepsy. Journal of Placantonakis D, Shariff S, Lafaille F, Labar D, Harden C,
Neurosurgery 1998;89:177-82. Hosain S, et al. Bilateral intracranial electrodes for lateralizing
intractable epilepsy: efficacy, risk and outcome. Neurosurgery
McClelland 2007 {published data only} 2010;66:274-83.
McClelland S 3rd, Curran CC, Davey CS, Okuyemi KS. Intractable
pediatric temporal lobe epilepsy in the United States:
Informed decisions.

Rocamora 2009 {published data only} early predictive elements and outcome. Child's Nervous System
Rocamora R, Mader I, Zentner J, Schulze-Bonhage A. 2008;24(12):1437-45.
Epilepsy surgery in patients with multiple cerebral cavernous
Upchurch 2010 {published data only}
malformations. Seizure 2009;18(4):241-5.
Upchurch K, Stern JM, Salamon N, Dewar S, Engel J Jr,
Roth 2011 {published data only} Vinters HV, et al. Epileptogenic temporal cavernous
Roth J, Olasunkanmi A, MacAllister WS, Weil E, Uy CC, malformations: operative strategies and postoperative seizure
Devinsky O, et al. Quality of life following epilepsy surgery for outcomes. Seizure 2010;19(2):120-8.
children with tuberous sclerosis complex. Epilepsy & Behavior
Vachrajani 2012 {published data only}
2011;20(3):561-5.
Vachhrajani S, de Ribaupierre S, Otsubo H, Ochi A, Weiss SK,
Sakuta 2005 {published data only} Donner EJ, et al. Neurosurgical management of frontal lobe
Sakuta R, Otsubo H, Nolan MA, Weiss SK, Hawkins C, Rutka JT, epilepsy in children. Journal of Neurosurgery. Pediatrics
et al. Recurrent intractable seizures in children with cortical 2012;10(3):206-16.
dysplasia adjacent to dysembryoplastic neuroepithelial tumor.
Vadera 2012 {published data only}
Journal of Child Neurology 2005;20(4):377-84.
Vadera S, Kshettry VR, Klaas P, Bingaman W. Seizure-free
Schwartz 2006 {published data only} and neuropsychological outcomes after temporal lobectomy
Schwartz TH, Jeha L, Tanner A, Bingaman W, Sperling MR. Late with amygdalohippocampectomy in pediatric patients with
seizures in patients initially seizure free after epilepsy surgery. hippocampal sclerosis. Journal of Neurosurgery. Pediatrics
Epilepsia 2006;47(3):567-73. 2012;10(2):103-7.
Smyth 2007 {published data only} Vadlamudi 2004 {published data only}
Smyth MD, Limbrick DD Jr, Ojemann JG, Zempel J, Robinson S, Vadlamudi L, So EL, Worrell GA, Mosewich RK, Cascino GD,
O'Brien DF, et al. Outcome following surgery for temporal lobe Meyer FB, et al. Factors underlying scalp-EEG interictal
epilepsy with hippocampal involvement in preadolescent epileptiform discharges in intractable frontal lobe epilepsy.
children: emphasis on mesial temporal sclerosis. Journal of Epileptic Disorders 2004;6(2):89-95.
Neurosurgery 2007;106(3 Suppl):205-10.
Weinand 2001 {published data only}
Soeder 2009 {published data only} Weinand ME, Kester MM, Labiner DM, Ahern GL. Time from
Soeder BM, Gleissner U, Urbach H, Clusmann H, Elger CE, ictal subdural EEG seizure onset to clinical seizure onset:
Vincent A, et al. Causes, presentation and outcome of lesional prognostic value for selecting temporal lobectomy candidates.
adult onset mediotemporal lobe epilepsy. Journal of Neurology, Neurological Research 2001;23:599-604.
Wetjen 2006 {published data only}
Stavem 2005 {published data only} Wetjen NM, Cascino GD, Fessler AJ, So EL, Buchhalter JR,
Stavem K, Guldvog B. Long-term survival after epilepsy surgery Mullan BP, et al. Subtraction ictal single-photon emission
compared with matched epilepsy controls and the general computed tomography coregistered to magnetic resonance
population. Epilepsy Research 2005;63(1):67-75. imaging in evaluating the need for repeated epilepsy surgery.
Journal of Neurosurgery 2006;105(1):71-6.
Stavem 2008 {published data only}
Wetjen 2009 {published data only}
Stavem K, Bjornaes H, Langmoen IA. Long-term seizures and
quality of life after epilepsy surgery compared with matched Wetjen NM, Marsh WR, Meyer FB, Cascino GD, So E, Britton JW,
controls. Neurosurgery 2008;62(2):326-34. et al. Intracranial electroencephalography seizure onset
patterns and surgical outcomes in nonlesional extratemporal
Stefan 2004 {published data only} epilepsy. Journal of Neurosurgery 2009;110(6):1147-52.
Stefan H, Walter J, Kerling F, Blumcke I, Buchfelder M.
Wieser 2003a {published data only}
Supratentorial cavernoma and epileptic seizures. Are there
predictors for postoperative seizure control? [Supratentorielle Wieser HG, Ortega M, Friedman A, Yonekawa Y. Long-term
Kavernome und epileptische Anfalle. Gibt es Pradiktoren fur seizure outcomes following amygdalohippocampectomy.
postoperative Anfallskontrolle?]. Nervenarzt 2004;75(8):755-62. Journal of Neurosurgery 2003;98(4):751-63.
Stefan 2008 {published data only} Wieser 2003b {published data only}
Stefan H, Hopfengartner R, Kreiselmeyer G, Weigel D, Rampp S, Wieser HG, Hane A. Antiepileptic drug treatment before and
Kerling F, et al. Interictal triple ECoG characteristics of temporal after selective amygdalohippocampectomy. Epilepsy Research
lobe epilepsies: an intraoperative ECoG analysis correlated with 2003;55(3):211-23.
surgical outcome. Clinical Neurophysiology 2008;119(3):642-52.
Yasuda 2010a {published data only}
Teutonico 2008 {published data only} Yasuda CL, Costa AL, Franca M Jr, Pereira FR, Tedeschi H,
Teutonico F, Mai R, Devinsky O, Lo Russo G, Weiner HL, de Oliveira E, et al. Postcraniotomy temporalis muscle atrophy:
Borrelli P, et al. Epilepsy surgery in tuberous sclerosis complex: a clinical, magnetic resonance imaging volumetry and

Informed decisions.

electromyographic investigation. Journal of Orofacial Pain seizures in temporal lobectomy patients. Epilepsy Research
2010;24(4):391-7. 1996;24:119-26.
Yasuda 2010b {published data only} Engel 1987

Yasuda CL, Morita ME, Alessio A, Pereira AR, Balthazar ML, Engel Jr J. Approaches to localization of the epileptogenic
Saude AV, et al. Relationship between environmental factors lesion. Surgical Treatment of the Epilepsies. New York: Raven
and gray matter atrophy in refractory MTLE. Neurology Press, 1987:75-96.
2010;74(13):1062-8.
Engel 1993a
Zupanc 2010 {published data only} Engel JJ, Van Ness PC, Rasmussen TB, Ojemann LM. Outcome
Zupanc M, dos Santos Rubio E, Werner R, Schwabe M, with respect to epileptic seizures. In: Engel J editor(s). Surgical
Mueller W, Lew S, et al. Epilepsy surgery outcomes: quality of Treatment of the Epilepsies. 2nd Edition. New York: Raven
life and seizure control. Pediatric Neurology 2010;42:12-20. Press, 1993:609-21.
Engel 1993b
Additional references Engel Jr J, Shewmon DA. Overview. Who should be considered
ACROBAT 2014 a surgical candidate?. Surgical Treatment of the Epilepsies. 2nd
Edition. New York: Raven Press, 1993:23-34.
Sterne JAC, Higgins JPT, Reeves BC on behalf of the
development group for ACROBAT-NRSI. A Cochrane Risk Of Bias Engel 1996
Assessment Tool: for Non-Randomized Studies of Interventions
Engel Jr J. Surgery for seizures. New England Journal of
(ACROBAT-NRSI). Available from http://www.riskofbias.info
Medicine 1996;334:647-53.
(accessed 16 February 2015) 24 September 2014; Vol. Version
1.0.0. Engel 2003
Annegers 1979 Engel Jr J, Wiebe S, French J, Sperling M, Williamson P,
Spencer D, et al. Practice parameter: temporal lobe and
Annegers J, Hauser WA, Elveback LR. Remission of seizures and
localized neocortical resections for epilepsy. Report of the
relapse in patients with epilepsy. Epilepsia 1979;20:729-37.
Quality Standards Subcommittee of the American Academy
Berg 1998 of Neurology, in association with the American Epilepsy
Society and the American Association of Neurological Surgeon.
Berg AT, Walczak T, Hirsch LJ, Spencer SS. Multivariable
Neurology 2003;60:538-47.
prediction of seizure outcome one year after resective epilepsy
surgery: development of a model with independent validation. Feindel 2009
Epilepsy Research 1998;29:185-94.
Feindel W, Leblanc R, de Almeida A. Epilepsy surgery: historical
Cleary 2012 highlights 1909–2009. Epilepsia 2009;50(Suppl 3):131-51.
Cleary RA, Thompson PJ, Fox Z, Foong J. Predictors of Fisher 2014
psychiatric and seizure outcome following temporal lobe
Fisher RS, Acevedo C, Arzimanoglou A, Bogacz A, Cross JH,
epilepsy surgery. Epilepsia 2012;53(10):1705-12.
Elger CE, et al. ILAE Official Report: A practical clinical definition
Cockerell 1995 of epilepsy. Epilepsia 2014;55(4):475-82.
Cockerell OC, Johnson AL, Sander JWAS, Hart YM, Shorvon SD. Guyatt 2008
Remission of epilepsy: results from the National General
Guyatt GH, Oxman AD, Vist G, Kunz R, Falck-Ytter Y, Alonso-
Practice Study of Epilepsy. Lancet 1995;346:140-4.
Coello P, et al. GRADE: an emerging consensus on rating
Collaborative 1992 quality of evidence and strength of recommendations. BMJ
2008;336(7650):924-6.
Collaborative Group for the Study of Epilepsy. Prognosis of
epilepsy in newly referred patients. A multicenter prospective Hayden 2006
study of the effects of monotherapy on the long-term course of
Hayden JA, Côté P, Bombardier C. Evaluation of the quality
epilepsy. Epilepsia 1992;33:45-51.
of prognosis studies in systematic reviews. Annals of Internal
Costa 2011 Medicine 2006;144(6):427-37.
Costa J, Fareleira F, Ascencao R, Borges M, Sampaio C, Vaz- Higgins 2011
Carneiro A. Clinical comparability of the new antiepileptic drugs
Higgins JPT, Green S (editors). Cochrane Handbook for
in refractory partial epilepsy: a systematic review and meta-
Systematic Reviews of Interventions Version 5.1.0 [updated
analysis. Epilepsia 2011;52:1280-91.
March 2011]. The Cochrane Collaboration, 2011. Available from
Davis 1996 www.cochrane-handbook.org.
Davies KG, Hermann BP, Dohan Jr FC, Foley KT, Bush AJ, ILAE 2001
Wyler AR. Relationship of hippocampal sclerosis to
Wieser HG, Blume WT, Fish D, Goldensohn E, Hufnagel A, King D,
duration and age of onset of epilepsy, and childhood febrile
et al. Commission on Neurosurgery of the International League

Informed decisions.

Against Epilepsy (ILAE). ILAE Commission Report. Proposal for a Sarkisian 1999
new classification of outcome with respect to epileptic seizures Sarkisian MR, Holmes GL, Carmant L, Liu Z, Yang Y, Stafstrom CE.
following epilepsy surgery. Epilepsia 2001;42:282-6. Effects of hyperthermia and continuous hippocampal
stimulation on the immature and adult brain. Brian and
Kirkham 2010
Development 1999;21:318-25.
Kirkham JJ, Dwan KM, Altman DG, Gamble C, Dodd S, Smyth R,
et al. The impact of outcome reporting bias in randomised Scott 2002
controlled trials on a cohort of systematic reviews. BMJ Scott RC, Gadian DG, King MD, Chong WK, Cox TC, Neville BG, et
2010;340:c365. al. Magnetic resonance imaging findings within 5 days of status
epilepticus in childhood. Brain 2002;125:1951–9.
Kwan 2000
Kwan P, Brodie MJ. Early identification of refractory epilepsy. Szabo 1999
New England Journal of Medicine 2000;342(5):314-9. Szabo CA, Wyllie E, Siavalas EL, Najm I, Ruggieri P, Kotagal P,
et al. Hippocampal volumetry in children 6 years or younger:
Kwan 2010b
assessment of children with and without complex febrile
Kwan P, Arzimanoglou A, Berg AT, Brodie MJ, Allen Hauser W, seizures. Epilepsy Research 1999;33:1-9.
Mathern G, et al. Definition of drug resistant epilepsy: consensus
proposal by the ad hoc Task Force of the ILAE Commission on Thom 2009
Therapeutic Strategies. Epilepsia 2010;51(9):1922. Thom M, Eriksson S, Martinian L, Caboclo LO, McEvoy AW,
Duncan JS, et al. Temporal lobe sclerosis associated
Lefebvre 2011
with hippocampal sclerosis in temporal lobe epilepsy:
Lefebvre C, Manheimer E, Glanville J. Chapter 6: Searching for neuropathological features. Journal of Neuropathology and
studies. In: Higgins JPT, Green S (editors). Cochrane Handbook Experimental Neurology 2009;68(8):928-38.
for Systematic Reviews of Interventions Version 5.1.0 (updated
March 2011). The Cochrane Collaboration, 2011. Available from Tonini 1997
www.cochrane-handbook.org. Tonini C, Beghi E, Vitezic D. Prognosis following epilepsy
surgery: a systematic review. Neurology Research International
Maher 1995
1997;2:1-6.
Maher J, McLachlan, RS. Febrile convulsions. Is seizure duration
the most important predictor of temporal lobe epilepsy?. Brain
1995;118(Pt 6):1521-8. References to other published versions of this review
Mann 1996 Tonini 2004
Mann T (1996). Clinical Guidelines: Using Clinical Guidelines to Tonini C, Beghi E, Berg AT, Bogliun G, Giordano L, Newton RW,
Improve Patient Care Within the NHS. London: Department of et al. Predictors of epilepsy surgery outcome: a meta-analysis.
Health.. Epilepsy Research 2004;62:75-87.
McIntosh 2001 West 2013

McIntosh AM, Wilson SJ, Berkovic SF. Seizure outcome after West S, Nolan SJ, Newton J, Sudan A, Pulman J, Newton R,
temporal lobectomy: current research practice and findings. Gandhi S. Surgery for epilepsy. Cochrane Database of Systematic
Epilepsia 2001;42:1288-307. Reviews 2013, Issue 6. [DOI: 10.1002/14651858.CD010541]
National 1990a
* Indicates the major publication for the study
National Institutes of Health Consensus Conference. Surgery for
epilepsy. JAMA 1990;264:729-33.

CHARACTERISTICS OF STUDIES
Characteristics of included studies [ordered by study ID]

Aaberg 2012
Methods Retrospective case series
Participants 54 children from a single Norwegian centre
Interventions Lesionectomy or hemispherotomy; 44 had resective surgery, temporal and extratemporal
Outcomes Assessed as Engel class at 12 and 24 months post-surgery

Informed decisions.

Aaberg 2012 (Continued)
Notes —

Adam 1996
Methods Prospective case series
Participants 30 adults aged 18 to 44 from single French centre
Interventions Antero-medial temporal lobe resections
Outcomes Seizure outcome assessed by Engel class at a mean of 24 months and at least 12 months after surgery
Notes —

Adelson 1992
Participants 33 children aged less than 15 years at 1 American centre
Interventions Temporal or extended temporal lobe resections
Outcomes Seizure outcome assessed by the presence of seizure freedom at least 18 months after surgery
Notes —

Alfstad 2011
Participants 48 patients, adults and children less than 14 years old, who had had surgery for epilepsy
Interventions Type of surgery: temporal and extratemporal
Outcomes Seizure outcome assessed at 2 years post-surgery using Engel's classification
Notes —

Althausen 2013
Participants 61 children (aged 6 years and above) and adults from single German centre
Interventions Hemispherectomy
Outcomes Reported as seizure-free at least 12 months following surgery

Informed decisions.

Althausen 2013 (Continued)
Notes —

Arruda 1996
Participants 74 adults from a single Canadian centre
Interventions Selective amygdalohippocampectomy or anterior temporal lobe resection
Outcomes Seizure outcome assessed by Engel class at a mean of 33.1 months
Notes Study of people with non-lesional temporal lobe epilepsy who had MRI volumetric studies

Awad 1991
Participants 47 children and adults from a single American centre
Interventions Surgical objectives included biopsy of the structural lesion and maximum resection of the lesion as de-
fined on neuroimaging studies and of the epileptogenic area whenever possible. Lesions were tempo-
ral and extratemporal
Outcomes Seizure outcome assessed as those seizure-free at 12 months
Notes —

Babini 2013
Participants 30 participants, age range 3 to 18 years, who underwent surgery for histopathologically confirmed low-
grade tumours, in which seizures were the only clinical manifestation, from a single Italian centre
Interventions Lesionectomy or tailored lesionectomy (i.e. tumour plus neighbouring epileptogenic region)
Outcomes Assessed as Engel class at least 12 months post-surgery
Notes —

Battaglia 2006
Methods Case series; unclear whether prospective or retrospective
Participants 45 children with refractory epilepsy operated on before 7 years from 1 Italian centre

Informed decisions.

Battaglia 2006 (Continued)
Interventions 19 had hemispherectomy, 9 had multilobar resection, 17 had focal resection - 9 temporal, 4 frontal, 4
parieto-occipital
Outcomes Seizure outcome after at least 2 years from time of surgery
Notes Focus on neurocognitive outcome

Baumann 2007
Methods Multicentre retrospective case series
Participants Participants were 168 consecutive children and adults with a single supratentorial cerebral cavernous
malformation and symptomatic epilepsy. Centres from Switzerland, Italy, Germany, USA and Canada
Outcomes Seizure outcome by Engel class was determined in the first, second and third postoperative years
Notes —

Bautista 2003
Participants 55 patients aged 17 to 57 years with a histopathological diagnosis of focal cortical dysplasia
Interventions Type of surgery - temporal and extratemporal
Outcomes Seizure outcome after at least 12 months of follow-up from time of surgery
Notes —

Bell 2009
Participants 44 patients, age range 13 to 62 years with a non-lesional modern "seizure protocol" MRI who under-
went anterior temporal lobectomy for treatment of medically refractory partial epilepsy
Interventions Temporal lobe surgery
Outcomes Seizure outcome measured using Engel's classification at least 1 year post-surgery
Notes Study of people with "non-lesional MRIs"

Benifla 2006

Informed decisions.

Benifla 2006 (Continued)
Participants 126 children who had surgery at 1 Canadian centre over a 10-year period
Interventions Anterior temporal lobectomy (ATL) with resection of mesiotemporal structures. The resection margin
extended 5 cm to 6 cm from the temporal pole of the nondominant hemisphere (and included the su-
perior temporal gyrus), and 4 cm to 5 cm in the dominant hemisphere, modified according to the pre-
sentation, imaging findings and lesion localisation. Patients with dorsal temporal or basal temporal le-
sions underwent lesionectomy or ATL without removal of the mesiotemporal structures. Mesiotempo-
ral structures were removed in certain cases, particularly if the lesions impinged upon or involved the
hippocampus or amygdala. Intraoperative ECoG was performed in 94 patients
Outcomes Seizure outcome assessed using Engel's classification at follow-up at least 24 months after surgery
Notes —

Berkovic 1995
Participants 135 children and adults from a single Australian centre
Interventions Standard anterior temporal lobectomy, including partial hippocampectomy, for dominant temporal
lobe removals, 3.5 cm of the lateral temporal lobe was excised; for nondominant removals, 5.0 cm was
excised. The hippocampus was excised microsurgically, usually to the level of the posterior midbrain.
Foreign tissue lesions were completely excised where possible, and the anterior 2 cm of hippocampus
was also removed. When the foreign tissue lesion was located in the lateral temporal region, the hip-
pocampus was not resected unless it appeared abnormal on MRI
Outcomes Seizure outcome assessed as Engel class at last follow-up (after at least 18 months follow-up) or at least
2 years of seizure-free remission at last follow-up
Notes —

Blount 2004
Participants 30 consecutive Canadian children from 1 centre followed for at least 30 months
Interventions Multiple subpial transections
Outcomes Seizure outcome by Engel class determined at mean follow-up of 3.5 years after surgery
Notes —

Blume 2004

Informed decisions.

Blume 2004 (Continued)
Participants 70 participants with intractable focal epilepsy and no specific lesion, as determined by both MRI (mag-
netic resonance imaging) and histopathology; age ranged from 6 to 65 years with a mean age of 31
years.
Interventions Temporal and extratemporal
Outcomes Seizure outcome measured using Engel's classification at least 2 years post-surgery
Notes A study of non-lesional intractable epilepsy

Boesebeck 2007
Participants 81 patients, aged 16 to 53 years at surgery, with lesional focal epilepsies of the extratemporal cortex
with resistance to at least 2 antiepileptic drugs, from a single German centre
Interventions Extratemporal surgery
Outcomes Seizure outcome measured using Engel's classification at 2-year follow-up post-surgery
Notes —

Boshuisen 2010
Participants 43 children who had hemispherectomy for intractable hemispheric epilepsy from a single Dutch centre
Interventions Functional hemispherectomy
Outcomes Seizure outcome by Engel class determined at least 12 months after surgery
Notes —

Brainer-Lima 1996
Participants 32 children and adults from a single Brazilian centre with a tumour and intractable epilepsy
Interventions Temporal corticectomy with amygdalo-hippocampectomy, temporal corticectomy, extratemporal cor-

ticectomy, posterior hippocampectomy and lesionectomy with stereotactic guidance
Outcomes Seizure outcome assessed by Engel class at a mean of 26.3 months
Notes —

Informed decisions.

Britton 1994
Participants 51 children and adults from a single American centre with a tumour and intractable epilepsy
Interventions Lesionectomy or lesion resection and corticectomy
Outcomes Seizure outcome assessed by Engel class at least 24 months postoperatively
Notes —

Caraballo 2011
Participants 45 children, aged 2 months to 18 years with a medically refractory epilepsy and hemispheric lesions
from a single Argentinian centre
Interventions Extratemporal
Outcomes Seizure freedom assessed using Engel's classification at least 1 year post-surgery
Notes —

Cascino 1995
Participants 165 children and adults from a single American centre with intractable temporal lobe epilepsy
Notes —

Chabardes 2005
Participants 48 consecutive adults from France and Italy with drug-refractory temporal lobe epilepsy
Interventions All the patients underwent tailored anterior temporal lobectomy that included the temporal pole, the
hippocampus, the parahippocampal gyrus and the anterior part of the lateral temporal cortex
Outcomes Seizure outcome by Engel class determined at least 48 months after surgery and related to site relative
to pole of temporal lobe seizure onset
Notes —

Informed decisions.


Chang 2009
Participants Results presented on 57 American participants with cavernomas and epilepsy out of a total group of
164 participants
Interventions Microsurgical resection of supratentorial cerebral cavernous malformations
Outcomes Seizure outcome by Engel class determined at 12 months after surgery
Notes —

Chee 1993
Participants 40 adults from a single American centre with temporal lobe epilepsy
Interventions Temporal lobectomy
Outcomes Seizure outcome assessed by seizure freedom at least 12 months postoperatively
Notes This was a study of non-lesional epilepsy with FDG-PET scanning

Chkhenkeli 2007
Participants 129 adults and children from Georgia and the USA with bitemporal epileptiform abnormalities in multi-
ple scalp EEGs
Interventions Temporal lobectomies were performed in 85 of 129. The temporal lobe resections included 2 modifi-
cations of the surgery, depending on the hemispheric dominance. The "standard temporal lobe resec-
tion" in "en block" modification was performed in the non-dominant hemispheres (29/67 patients).
This resection usually included 6.0 cm to 6.5 cm of lateral cortex, uncus, amygdala and 2 cm to 4 cm of
the anterior hippocampus. In the dominant hemisphere (38/67 patients) the extension of cortical re-
section was reduced to 3 cm to 4 cm, and usually performed as a modification named "anterior medi-
al temporal lobectomy". This modification includes incision of the temporal lobe cortex for 3 cm to 3.5
cm from the temporal pole along the inferior surface of the superior temporal gyrus
Notes —

Choi 2004a
Participants 35 Korean teenagers and adults with temporal lobe epilepsy associated with tumour

Informed decisions.

Choi 2004a (Continued)
Interventions Resection of epileptogenic area and tumour (guided by ECoG as required)
Outcomes Seizure outcome by Engel class determined at mean of 33 months and at least 15 months after surgery
Notes —

Chung 2005
Participants 128 adults and children with epilepsy secondary to cortical dysplasia diagnosed on histology postoper-
atively from single Korean centre
Notes —

Cossu 2005
Participants 174 children and adults from 1 Canadian centre operated on over a 7-year period from a single Italian
centre
Interventions Temporal and extratemporal surgery: corticectomy, n = 58; corticectomy and lesionectomy, n = 112;
and lesionectomy, n = 3
Notes —

Cossu 2008
Participants 113 children from 1 Italian centre
Interventions 72 had a complete lesionectomy. Resection sites were as follows: 43 temporal, 32 frontal, 20 posterior,
9 including central, 4 temporal plus and 5 wide multilobar
Notes —


Informed decisions.

Costello 2009
Participants 42 with disabling, medically refractory focal epilepsy operated on at 1 American centre
Interventions 11 participants underwent a standard left anterior-medial temporal lobectomy (including amygdalo-
hippocampectomy); 17 a standard right anterior-medial temporal lobectomy; 3 a limited left frontal re-
section; 3 a limited right frontal resection and 2 a left temporal lesionectomy (without removal of the
amygdalohippocampal complex). The following operations were performed in single patients: right
temporal neocortical resection of extensive malformation of cortical development, right posterior tem-
poral resection, left posterior temporal resection, right temporal lesionectomy, corpus callosotomy
and left parietal lesionectomy. Intra-operative mapping of language was performed in 2
Outcomes Seizure outcome was assessed using Engel's classification at between 1 and 14.4 years postoperatively
Notes Included only those 45 years or older

Cukiert 2002
Participants 100 adults from single Brazilian centre
Interventions Corticoamygdalohippocampectomy on the side of the mesial temporal sclerosis, consisting of cortical
resection, including the superior, middle and inferior temporal lobes, the fusiform and parahippocam-
pal gyri, with its posterior border at the level of the central artery, total hippocampectomy, and resec-
tion of the intratemporal portion of the amygdala. The central artery was used as the landmark for the
posterior border of the cortical resection (a proportional method), instead of measuring distances from
the tip of the temporal lobe (a quantitative method).
Outcomes Seizure outcome assessed using Engel's classification at follow-up from 18 to 48 months after surgery
Notes —

Dagar 2011
Participants 118 children, aged 0.3 to 18 years at the time of surgery, with a medically refractory epilepsy, from a
single Indian centre
Outcomes Seizure outcome as measured using Engel's classification at least 12 months after surgery
Notes —

Dalmagro 2005

Informed decisions.

Dalmagro 2005 (Continued)
Participants 44 children and adults with posterior cortex epilepsy from a single Brazilian centre
Interventions Multilobar resection, lesionectomy and lobectomy in five (11.63%). According to PO MRI, surgeries were
considered complete in 16 (37.21%) and incomplete in 27 (62.79%) of 43 patients. However, the com-
pleteness of the resection had no influence on surgical outcome
Outcomes Assessed as Engel class at 12 months post-surgery
Notes —

de Tisi 2011
Methods Retrospective cohort study
Participants 615 people, aged 16 to 63 years at surgery, who had undergone surgery for epilepsy, from a single UK
centre
Interventions Temporal and extratemporal surgery
Outcomes Seizure freedom assessed using ILAE outcome score at follow-up, at least 12 months from surgery
Notes —

Delbeke 1996
Participants 38 participants aged 15 to 59 years from single American centre with temporal lobe epilepsy who had
FDG-PET as part of their pre-operative assessment
Interventions Lesionectomy with or without neocortical resection
Outcomes Seizure outcome assessed by Engel class at least 18 months postoperatively
Notes —

Dellabadia 2002
Participants Initially 99 children and adults admitted for surgical selection with a temporal or extratemporal lobe
epilepsy from an American centre
Interventions Of the 69 patients evaluated, 35 had a focal resection (33 temporal, 2 frontal)
Notes Focus on which pre-operative tests are most discriminatory of good outcome

Informed decisions.


Devlin 2003
Participants 33 children who underwent hemispherectomy at a single UK centre between 1991 and 1997
Interventions Functional hemispherectomy involving a modified approach with a limited suprasylvian window but a
large temporal lobectomy. The insular cortex was undercut
Notes —

Donadio 2011
Participants 110 participants, children and adults ranging in age from 1 year to 52 years, with a drug-resistant
epilepsy
Interventions Extratemporal and temporal surgery including lesionectomies, lobectomies, callostomies, multiple
subpial transections, hemispherectomies and insertion of vagal nerve stimulators
Outcomes Assessed using Engel's classification at follow-up at least 12 months after surgery
Notes —

Dorward 2011
Participants 33 children, aged 3 to 19 years at surgery, with an intractable epilepsy, no lesion on MRI scan and who
had invasive EEG monitoring with subdural grid/strip electrodes
Interventions Extratemporal resections or multiple subpial resections
Notes —

Duchowny 1998
Participants 31 young children from a single American centre
Interventions Temporal and extratemporal lesionectomy
Notes —

Informed decisions.


Dunkley 2011
Participants 42 children under 36 months of age who had surgery for epilepsy
Interventions Temporal and extratemporal surgery and hemispherectomies
Outcomes Assessed using Engel's classification at follow-up at least 12 months after surgery
Notes —

Dunlea 2010
Participants 199 Irish participants with at least 1-year follow-up who underwent resective surgery for refractory
epilepsy since 1975
Interventions Interventions included anterior temporal lobectomy; amygdalo-hippocampectomy; neocorticectomy;

lesionectomy and frontal lobe resection
Outcomes Engel's criteria were used to classify seizure outcome at 1, 2, 5, 10, 15 and > 15 years follow-up
Notes —

Elsharkawy 2008a
Participants 218 consecutive adults with extratemporal lesions who underwent resective surgical treatment for in-
tractable focal epilepsy from a single German centre between 1991 and 2005
Interventions Resection of epileptogenic zone: frontal lobe 95; posterior cortical 103 and multilobar 20
Notes —

Elsharkawy 2009a
Participants 434 German adults from 1 centre
Interventions 1) Anterior temporal lobe resection included the pole of the temporal lobe. The laterodorsal resection
line was delineated by EEG as well as abnormalities noted by MRI. The size of the resection ws 2.5 cm to
4 cm in the language-dominant hemisphere and 3 cm to 6 cm in the non-dominant hemisphere. Part of
the procedure was to remove the parahippocampal gyrus, hippocampus and amygdala

Informed decisions.

Elsharkawy 2009a (Continued)
2) Apical temporal resection: tailored resection of the lesion in the apex of the temporal lobe with amy-
dalectomy, and maximal 4 cm laterodorsal cortex from the pole, extension of the resection was guided
by ECoG
3) Temporal lesionectomy included only a singular lesion resection as defined by EEG and MR imaging,
but saved the eloquent cortex. In the case of dual pathology a lesionectomy and a selective amygdalo-
hippocampectomy were performed, and the dorsal resection was guided by means of intraoperative
ECoG
4) Selective amygdalohippocampectomy included only a resection of the hippocampus or mesial
structures based on MRI and intraoperative findings (5 people only)
Outcomes Seizure outcome by Engel class determined at 24 months, 5, 10 and 16 years after surgery
Notes —

Elsharkawy 2011a
Participants 61 patients ranging in age from 5 to 58 years at the time of surgery with a refractory temporal lobe
epilepsy, MRI showing a lesion in the apex of the temporal lobe but normal hippocampus and intact
memory function
Interventions All had apical temporal lobe resections
Outcomes Seizure outcome measured at 2 years and 5 years follow-up post-surgery, measured according to En-
gel's classification
Notes —

Engman 2004
Participants 54 patients
Outcomes Seizure outcome not main focus but did give numbers for those who were seizure-free at 2 years fol-
low-up
Notes —

Erba 1992
Participants 46 children and adults from a single American centre
Interventions Standard or modified en bloc anterior temporal lobectomy. Depending on hemispheric dominance,
4 cm to 7 cm of lateral cortex of the temporal lobe was removed. Mesial structures (uncus, amygdala,

Informed decisions.

Erba 1992 (Continued)
hippocampus and hippocampal gyrus) were not removed when assessment suggested bilateral in-
volvement
Notes —

Erickson 2005
Participants 84 military beneficiaries at the only US military medical centre with a comprehensive epilepsy surgery
programme
Interventions Standard temporal lobectomy in the majority including mesial temporal lobe structures; margins ex-
tended posteriorally 3.5 cm to 4.0 cm in the dominant lobe and 5 cm to 6 cm in the non-dominant
lobes; 8 had amygdalohippocampectomies
Notes —

Fauser 2004
Participants 67 patients, aged 2 to 66 years, with histologically proven focal cortical dysplasias
Interventions Temporal lobe
Outcomes Seizure outcome measured using Engel's classification at least 12 months post-surgery, with a mean
follow-up period of 21.9 months
Notes —

Fujiwara 2012
Participants 44 operated after intracranial EEG over a 15-month period in 2 American centres
Interventions Lesionectomies - temporal and extratemporal
Outcomes Reported as seizure-free at least 12 months following surgery
Notes Focus was outcome related retrospectively to presence of high-frequency oscillations on ICEEG


Informed decisions.

Garcia 1991
Participants 55 children and adults from a single American centre operated on over a 3-year period
Outcomes Seizure outcome assessed by seizure freedom at 12 months postoperatively and annually thereafter
Notes Focus on postoperative seizures and outcome

Garcia 1994
Participants 51 participants from a single American centre operated on over a 3-year period
Interventions Anterotemporal resection including the amygdala and hippocampus
Notes Focus on value of qualitative pre-operative MRI findings

Gelinas 2011
Methods Single-centre retrospective case series
Participants 67 children between the ages of 3 months and 16 years with recurrent seizures attributable to a dis-
crete lesion on neuroimaging
Interventions Site of surgery was temporal and extratemporal
Outcomes Seizure outcome was examined 1 year post-surgery and at subsequent follow-ups using Engel's classifi-
cation
Notes —

Georgakoulias 2008
Participants 50 adult patients, mean age 34 years, with medically intractable medial temporal lobe epilepsy from a
single UK centre
Interventions Temporal
Outcomes Seizure outcome measured using Engel's classification. Intermediate and long-term with mean fol-
low-up of 6.2 years
Notes —

Informed decisions.


Gilliam 1997a
Participants 78 children and adults from a single American centre with mesial-basal temporal lobe epilepsy
Interventions En bloc neocorticectomy of the anterior 4.5 cm to 5.5 cm of the temporal lobe, sparing the superior
temporal gyrus
Outcomes Seizure outcome assessed by seizure freedom at least 12 months after surgery
Notes Study focuses on concordance of MRI and EEG findings and outcome

Gilliam 1997b
Participants 33 children from a single American centre
Interventions Extratemporal and temporal lobe cortical resection on children with intractable epilepsy
Outcomes Seizure outcome assessed by ILAE classification at a mean follow-up period of 2.7 years
Notes —

Goldstein 1996
Participants 33 children from a single American centre
Interventions Temporal lobe cortical resection on children with intractable epilepsy
Outcomes Seizure outcome assessed by Engel classification at least 24 months post-surgery
Notes —

Greiner 2011
Methods Retrospective case series from 2 centres in the US
Participants 54 participants, aged 6 months to 40 years at the time of surgery, who had had a hemispherectomy
Interventions Extratemporal i.e. all had hemispherectomy
Notes —

Informed decisions.


Grivas 2006
Participants 52 patients older than 50 years were operated on for intractable mesial or combined mesiolateral TLE
from 1 German centre
Interventions Temporal lobe resections
Outcomes Seizure outcome measured using Engel's classification with a mean follow-up period of 33 months
Notes —

Gyimesi 2007
Methods Retrospective case series; a German/Hungarian collaboration
Participants 130 adult patients (no age range provided) who had undergone epilepsy surgery for intractable medial
temporal lobe epilepsy
Outcomes Seizure freedom at 24 months after surgery compared with not seizure-free
Notes —

Hader 2004
Methods Retrospective case series from 1 centre in Canada
Participants 39 children, aged 2 months to 18.5 years, at surgery with a medically intractable epilepsy and focal cor-
tical dysplasia on histology
Outcomes Seizure outcome measured using Engel's classification at least 1.5 years after surgery
Notes —

Hajek 2009
Participants 35 participants with mesial temporal lobe epilepsies who had had MR spectroscopy prior to surgery,
from 1 Czech centre
Outcomes Seizure outcome measured using Engel's classification measured at least 24 months after surgery
Notes —

Informed decisions.


Hallbook 2010
Participants 110 children, 18 years or younger, with severe refractory epilepsy from a single US centre
Interventions Functional hemispherectomy
Outcomes Seizure freedom at time of follow-up, which was at least 12 months
Notes —

Hamiwka 2005
Participants 38 children, age at surgery ranging from 6 months to 18 years (mean, 9.6 years), with malformations of
cortical development
Outcomes Seizure outcome at 2, 5 and 10 years measured using Engel's classification
Notes —

Hartley 2002
Methods Retrospective case series from a single UK centre
Participants 35 children (24 females, 11 males; mean age 9.6 years, age range 11 months to 18 years) with a partial
epilepsy who had had a SPECT scan prior to surgery
Interventions Temporal, extratemporal and hemispherectomy
Outcomes Seizure outcome measured using Engel's classification; the range of follow-up since surgery was 3 to 6
years (mean 4.8 years)
Notes —

Hartzfield 2008
Participants 57 patients operated on for post-traumatic medial temporal lobe epilepsy from a single US centre
Outcomes Seizure outcome measured using Engel's classification with a mean follow-up of 4.84 years and a range
of 0.5 to 9 years

Informed decisions.

Hartzfield 2008 (Continued)
Notes —

Hemb 2010
Methods Retrospective case review from a single US centre
Participants 192 children operated on before 1997 compared with 397 children operated on from 1998 to 2008, all
with a refractory epilepsy
Interventions Extratemporal and temporal and hemispherectomies
Notes —

Holmes 1997
Participants 44 teenagers and adults with bitemporal, independent, interictal epileptiform patterns from a single
American centre
Outcomes Seizure outcome assessed by seizure freedom at least 12 months post-surgery
Notes —

Holmes 2000
Participants 126 children and adults with medically intractable extratemporal epilepsy from a single American cen-
tre
Interventions Extratemporal cortical resections
Notes —

Jack 1992
Participants 50 teenagers and adults with medically intractable temporal lobe epilepsy from a single American cen-
tre

Informed decisions.

Jack 1992 (Continued)
Interventions Anterior temporal lobectomy
Notes —

Janszky 2003a
Methods Retrospective case series from a single German centre
Participants 133 patients (aged 16 to 59 years) with hippocampal sclerosis-associated temporal lobe epilepsy
Outcomes Seizure outcome at 2 years post-surgery (for 84 patients) measured using Engel's classification
Notes —

Janszky 2003b
Methods Retrospective case series from a single German centre
Participants 147 patients (range 16 to 59 years) with intractable medial temporal lobe epilepsy who underwent
presurgical evaluation including high-resolution MRI and video-EEG monitoring with seizure registra-
tion
Outcomes Seizure outcome at 2 years post-surgery measured using Engel's classification
Notes —

Jaramillo-Betancur 2009
Methods Retrospective case series; nested case-control study
Participants 89 teenagers and adults from Columbia
Interventions Temporal lobectomies
Outcomes Seizure outcome by Engel class determined at 12 and 24 months after surgery
Notes —

Jayakar 2008

Informed decisions.

Jayakar 2008 (Continued)
Participants 102 children with non-lesional intractable partial epilepsy
Outcomes Seizure outcome measured according to Engel's classification at 2 years, 5 years and 10 years post-
surgery
Notes —

Jayalakshmi 2011
Methods Retrospective case review
Participants 87 children with refractory partial epilepsy
Interventions Temporal, extratemporal and hemispherectomy
Notes —

Jeha 2006
Participants 371 patients who underwent anterior temporal lobectomy to treat pharmacoresistant epilepsy
Outcomes Seizure freedom versus not seizure-free assessed at least 1 year after surgery
Notes —

Jehi 2012
Methods Retrospective case series from a single US centre
Participants 312 patients ranging in age from 2.5 to 74 years with an intractable temporal lobe epilepsy
Interventions Temporal resections
Notes —

Jennum 1993

Informed decisions.

Jennum 1993 (Continued)
Participants 64 children and adults with medically intractable temporal and extratemporal lobe epilepsy from a sin-
gle Danish centre
Interventions Tailored temporal or extratemporal lobe cortical resection
Notes —

Jeong 1999
Participants 93 consecutive children and adults with medically intractable mesial temporal lobe epilepsy from a sin-
gle Korean centre
Notes —

Kan 2008
Participants 58 children with an intractable localised epilepsy
Outcomes Seizure outcome measured according to Engel's classification at least 1 year post-surgery
Notes —

Kang 2009
Participants 244 adult patients with a mean age at surgery of 35 years (range 18 to 68 years) with an intractable tem-
poral lobe epilepsy and all of whom had a BMI > 26, i.e. overweight, obese or morbidly obese
Outcomes Seizure outcome measured according to Engel's classification at least 1 year post-surgery
Notes —


Informed decisions.

Kanner 2009
Participants 100 patients with a mean age of 31.2 years with temporal lobe epilepsy (TLE)
Outcomes Seizure outcome at 2-year follow-up measured using Engel's classification
Notes —

Kilpatrick 1997
Participants 75 consecutive teenagers and adults with medically intractable temporal lobe epilepsy from 2 Aus-
tralian centres
Interventions Those with hippocampal sclerosis underwent a tailored anterior temporal lobectomy with en bloc exci-
sion of the neocortical structures followed by microsurgical resection of the amygdala and en bloc exci-
sion of the hippocampal formation and parahippocampal gyrus. On the non-dominant side this includ-
ed excision of 4 cm of the superior temporal gyrus and the middle temporal gyrus, and inferior tempo-
ral gyrus to the vein of Labbe or 5 cm to 6 cm; in dominant lobectomy, the superior temporal gyrus was
left intact, the middle temporal gyrus was excised for either 4 cm to 5 cm or to the vein of Labbe, and
the inferior temporal was excised for either 4.5 cm to 5.5 cm or to the vein of Labbe. Patients with for-
eign tissue lesions either had an anterior temporal lobectomy, lesionectomy or neocorticectomy
Outcomes Seizure outcome assessed by Engel class at least 12 months post-surgery
Notes —

Kim 2009
Methods Retrospective case series from a single Korean centre
Participants 166 patients with intractable epilepsy related to focal cortical dysplasia aged 3 to 51 years with a mean
age of 24.7 years
Outcomes Seizure outcome, i.e. freedom or not at follow-up; mean length of postoperative follow-up was 7.94
years
Notes —

Kim 2010a
Participants 177 participants, aged between 11 and 51 years, at time of surgery who had had resective surgery and
intracranial EEG monitoring

Informed decisions.

Kim 2010a (Continued)
Interventions Temporal and extratemporal resections
Notes —

Kim 2010b
Participants 40 patients, aged 4 to 51 years, with refractory epilepsy
Interventions Temporal and extratemporal surgery
Outcomes Seizure freedom (as opposed to not seizure-free) measured at least 2 years postoperatively
Notes —

Kloss 2002
Methods Retrospective and prospective case series
Participants 68 participants under the age of 18 years from 1 German centre
Interventions Types of resection included lesionectomy, lesion and corticectomy, lobectomy and multilobar resec-
tions
Outcomes Seizure outcome was assessed using Engel's classification at 2 years postoperatively
Notes —

Knowlton 2008
Methods Prospective case series from a single US centre
Participants 62 participants with a mean age at surgery of 26 years, minimum age of 1 year and maximum age of
60 years, who required intracranial electroencephalography (ICEEG) because epileptic focus not suffi-
ciently localised with scalp EEG and MRI
Outcomes Seizure outcome measured using Engel's classification at least 1 year post-surgery; range of follow-up
was 1.5 to 10.5 years with the mean time being 4.2 years and median being 3.5 years. Patients with < 1
year follow-up were excluded
Notes —


Informed decisions.

Kral 2007
Methods Retrospective cases series from a single German centre
Participants 49 patients with a mean age at surgery of 18 years, range 5 to 47 years with focal cortical dysplasia
Outcomes Seizure outcome measured using the ILAE classification with a mean follow-up of 8.1 (SD 4.5) years
Notes —

Krsek 2013
Participants 106 children from 1 American centre
Interventions Resection of epileptogenic zone - temporal and extratemporal
Outcomes Assessed as Engel class at 24 months post-surgery
Notes Focus was outcome related retrospectively to pre-operative SPECT findings

Kuzniecky 1993
Participants 34 children and adults with medically intractable temporal lobe epilepsy from 1 American centre
Outcomes Seizure outcome assessed by modified Engel class at least 12 months post-surgery
Notes Focus on MRI findings and presence of febrile seizures

Kwan 2010
Methods Retrospective cases series
Participants 41 Canadian children undergoing hemispherectomy
Interventions Hemidecortication compared with peri-insular hemispherotomy
Notes —


Informed decisions.

Lackmayer 2013
Participants 45 consecutive patients with medically refractory unilateral mesial temporal lobe seizures from a single
Austrian centre
Interventions Either selective amygdalohippocampectomy or anteriormedial temporal lobectomy
Outcomes Assessed as seizure freedom 1, 2 or 3 years post-surgery
Notes Focus of study was postoperative depression related to seizure outcome

Lee 2006
Methods Retrospective cases series from a single Korean centre
Participants 51 patients with a mean age at surgery of 31.4 years, ranging from 16 to 50 years, with pathologically
proven mesial temporal sclerosis
Outcomes Seizure outcome measured using Engel's classification at follow-up, which lasted at least 4 years
Notes —

Lee 2008
Methods Not stated
Participants 71 participants with frontal lobe epilepsy with a mean age at surgery of 26.2 years, ranging from 12 to
57 years, from a single Korean centre
Outcomes Seizure outcome at least 2 years after surgery with an unclassified scale
Notes —

Lee 2010a
Participants 52 participants with intractable temporal lobe epilepsy; 19 patients were classified as children (< or =
18 years old), and 33 patients were classified as adults (> 18 years old)
Outcomes Whether seizure-free or not at 2 years and 4 years post-surgery
Notes —

Informed decisions.


Lee 2011
Participants 40 Korean participants treated for lesional mesial temporal lobe epilepsy between 1993 and 2008
Interventions Intervention before 2006 was anterior temporal lobectomy and from 2006 selective lesionectomy via a
transsylvian-transcisternal approach
Notes —

Lei 2008
Participants 250 cases of epilepsy caused by cerebral schistosomiasis in patients aged 17 to 66 years (mean 32.8
years)
Outcomes Seizure outcome at follow-up 4 to 5 years after operation measured using Engel's classification
Notes —

Li 1997
Participants 51 children and adults with medically intractable temporal and extratemporal lobe epilepsy from a
Canadian, UK and American centre
Interventions Lesionectomy, lesionectomy plus corticectomy or lobectomy, corticectomy without removal of the le-
sion, selective amygdalo-hippocampectomy, selective amygdalohippocampectomy and lesionectomy
Notes —

Li 1999
Methods Case series; not clear whether retrospective or prospective
Participants 38 teenagers and adults with medically intractable epilepsy associated with hippocampal sclerosis and
an additional lesion, which could have been temporal or extratemporal from a Canadian, UK, American
and 2 Australian centres

Informed decisions.

Li 1999 (Continued)
Interventions Lesionectomy (removal of the extrahippocampal lesion); mesial temporal resection (removal of the at-
rophic hippocampus); and lesionectomy plus mesial temporal resection (removal of both the lesion
and the atrophic hippocampus)
Notes —

Liang 2010
Methods Prospective study "randomly allocated" to either treatment group
Participants 60 Chinese participants with temporal lobe epilepsy and mental retardation
Interventions Half had anterior temporal lobectomy, half anterior corpus callosotomy combined with anterior tem-
poral lobectomy
Notes A study to compare the outcome of the 2 surgical approaches

Liang 2012
Participants 206 children undergoing surgical resection for epilepsy between 2001 and 2007 from 4 Chinese centres
Interventions Lesion resection, epileptogenic zone resection, anterior temporal lobectomy (involving the resection of
3 cm to 3.5 cm of the neocortex and 2 cm to 2.5 cm of the mesial structure of the left anterior temporal
lobe, or 3.5 cm to 4.5 cm of the neocortex and 2 cm to 3 cm of the mesial structure of the right anterior
temporal lobe, and selective amygdalohippocampotomy using a trans-Sylvian approach
Outcomes Assessed as seizure freedom at least 12 months post-surgery
Notes —

Liava 2012
Participants 53 children and young adults with extratemporal lobe epilepsy from 2 Italian centres
Interventions Tailored resections with localisation: 5 frontomesial, 6 fronto-dorsolateral, 6 fronto-mesial + dorsolat-

eral, 2 fronto-orbital, 5 fronto-operculo-insular, 1 fronto-central, 4 frontocentro-parietal, 3 fronto-cen-
tro-temporal, 5 parietal, 2 parietotemporal, 1 centro-parietal, 1 occipital, 7 occipito-temporal, and 5
temporo-parieto-occipital. In 5 cases, only a partial excision of the EZ was performed, due to its func-
tional intersection with eloquent areas: 1 fronto-dorsolateral, 1 fronto-mesial + dorsolateral, 1 fron-
to-central, 1 parietal, and 1 temporo-parieto-occipital; the remaining resections were considered as
complete

Informed decisions.

Liava 2012 (Continued)
Notes —

Lopez-Gonzalez 2012
Participants 130 children, aged 1 to 18 years at surgery, who had temporal lobe surgery
Outcomes Seizure outcome measured using Engel's classification at 1, 2, 5 and 12 years post-surgery
Notes —

Lorenzo 1995
Participants 48 participants (age not specified) with medically intractable focal frontal lobe epilepsy from 1 Ameri-
can centre
Interventions Focal cortical resection, i.e. a partial or complete frontal lobectomy with or without an identified mass
lesion, and a stereotactic resection of a frontal lobe MRI-identified epileptogenic lesion
Notes —

Madhavan 2007
Methods Retrospective case series from a French, 3 American and 1 Canadian centre
Participants 70 patients with tuberous sclerosis complex and epilepsy, mean age at surgery of 9.9 years
Outcomes Seizure outcome at follow-up (time from surgery to evaluation was 5.2 (8.0) years) measured using En-
gel's classification
Notes —

Mani 2006
Methods Retrospective cases series from a single US centre
Participants 132 children (< 18 years), who had intractable epilepsy; mean age at surgery was 8.17 years

Informed decisions.

Mani 2006 (Continued)
Interventions Extratemporal cortical resection and hemispherectomy
Outcomes Seizure outcome at 12 and 24 months post-surgery assessed using Engel's classification
Notes —

Mathern 1999
Methods Retrospective case series from a multi-centre Australian surgery for epilepsy programme
Participants 198 children with medically intractable temporal and extratemporal lobe epilepsy from 1 American
centre
Interventions The most common procedures were hemispherectomies followed by lobar resections
Notes —

McIntosh 2012
Methods Retrospective case series from an Australian centre
Participants 81 patients, aged 4 to 60 years at time of surgery with 12 people being less than 16 years, who had ex-
tratemporal resection
Interventions Extratemporal resections
Outcomes Seizure freedom measured at least 2 years post-surgery
Notes —

Mihara 2004
Methods Retrospective case series from a Japanese centre
Participants 357 patients with a medically intractable epilepsy, mean age at surgery calculated as 24.7 years; tem-
poral lobe group: 25.5 years (range 2 to 55 years); extratemporal group: 21.8 years (range 2 to 40 years)
Outcomes Seizure outcome measured using Engel's classification at least 12 months post-surgery
Notes —

Miserocchi 2013
Methods

Informed decisions.

Miserocchi 2013 (Continued)
Participants
Interventions
Outcomes
Notes

Morino 2009
Participants 62 people with temporal lobe seizures operated on in a single Japanese centre
Interventions All participants underwent trans-Sylvian selective amygdalohippocampectomy
Outcomes Seizure freedom 12 months postoperatively
Notes Study to determine the effects of selective surgery on memory outcome

Morris 1998
Methods Retrospective case series from a US centre
Participants 38 children and adults with medically intractable epilepsy who had had a resection of a ganglioglioma
in the temporal or extratemporal lobe over a 9-year period from 1 American centre
Interventions Tumour resection
Notes —

O'Brien 1996
Participants 46 teenagers and adults with medically intractable temporal lobe epilepsy over a 9-year period from 1
Australian centre
Interventions Standard anterior temporal lobectomy (4.5 cm to 5.5 cm lateral resection) with en block removal of the
mesial temporal structures or lesionectomy
Notes —


Informed decisions.

O'Brien 2000
Participants 36 children who had peri-ictal and interictal SPECT studies and extratemporal resective epilepsy
surgery performed at 1 American centre between June 1993 and June 1997
Interventions Resection of epileptogenic zone
Notes A study of whether subtraction ictal SPECT coregistered with MRI (SISCOM) is predictive of outcome

Oertel 2005
Methods Prospective randomised study from a German centre
Participants 30 patients; mean age and range: waterjet group 35.5 (18 to 70), aspirator group 34.7 (20 to 57), mean
age for all patients calculated as 35.1, with an intractable epilepsy
Outcomes Seizure outcome with a mean follow-up period of 2.15 years (range 1 to 3.5 years), measured as either
seizure-free or not
Notes —

Paglioli 2006
Participants 161 consecutive Brazilian participants with MTLE/HS
Interventions Anterior temporal lobectomy or a selective amygdalohippocampectomy
Outcomes Seizure outcome by Engel class determined at 24 months after surgery onwards
Notes —

Paolicchi 2000
Participants 75 children with intractable temporal or extratemporal lobe epilepsy at 2 American centres
Interventions Cortical resections. None of the temporal resections were 'standard'; all included the anterior neocor-
tical and mesial limbic structures, tailored posteriorly according to EEG, lesional data, and location of
language cortex. For seizures that originated posteriorly, resection of the temporal convexity and basal
neocortex was extended further posteriorly, with the vein of Labbe being undercut if needed. Extratem-
poral resections consisted of complete removal of the lesion combined with corticectomy tailored to
the epileptogenic region. Anterior frontal epileptogenic regions were often treated by medial or lateral

Informed decisions.

Paolicchi 2000 (Continued)
wedge resections; posterior frontal, parietal and occipital foci were more likely to be treated by tailored
corticectomy alone
Notes —

Park 2002
Participants 148 participants younger than 18 years who underwent surgery for the relief of medically intractable
epilepsy, with a mean age at surgery of 13.4 years (range 5 months to 18 years)
Outcomes Seizure outcome measured using Engel's classification at least 12 months after surgery
Notes —

Park 2006
Participants 30 patients with cortical dysplasia (CD) and epilepsy range, age range 1.5 to 18.3 years
Outcomes Seizure outcome measured using Engel's classification with a mean follow-up period of 3.2 years and a
minimum of 12 months
Notes —

Perego 2009
Participants 37 adults from 1 Spanish centre
Interventions 25 participants underwent anteromedial temporal lobe resection; 5 underwent temporal complete
lobectomy; 6 a lesionectomy with well-demarcated lesions and 1 a selective transventricular amyg-
dalohippocampectomy
Outcomes Seizure outcome was assessed using Engel's classification at 1 and 3 years postoperatively
Notes —


Informed decisions.

Perry 2010
Participants 83 American participants less than 18 years of age with incomplete resection (defined by intraoperative
or extraoperative subdural EEG data and postoperative MRI where a lesion was present) for epilepsy,
with 2 years of follow-up
Interventions Lesional resection - excluding those who had corpus callosotomy, vagal nerve stimulator placement,
multiple subpial transections as their sole procedure, as were those who had an hemispherectomy
Notes Note: this is a follow-up study of those with original incomplete resection

Phi 2009
Participants 87 participants with tumour-related temporal lobe epilepsy from a single Korean centre
Interventions Temporal lobe lesionectomy with or without hippocampectomy
Notes —

Phi 2010
Participants 41 paediatric participants with focal cortical dysplasia from a single Korean centre
Interventions Temporal or extratemporal lobe surgery
Notes —

Pinheiro-Martins 2012
Methods Retrospective case series from a Brazilian centre
Participants 70 participants, aged 1 to 52 years at time of surgery, with a refractory frontal lobe epilepsy
Outcomes Seizure outcome measured according to Engel's classification at time of follow-up, which was at least 4
years after surgery
Notes —

Informed decisions.


Prevedello 2000
Participants 84 adults with intractable temporal lobe epilepsy at 1 German centre
Notes —

Raabe 2012
Methods Retrospective case series from a German centre
Participants 80 participants with a drug-resistant focal epilepsy and either a cavernous angioma or an arterio-ve-
nous malformation in underlying histology
Interventions Not stated whether temporal or extratemporal resections
Outcomes Seizure outcome measured using Engel's classification at follow-up at least 2 years post-surgery
Notes —

Radhakrishnan 1998
Methods A prospective and retrospective case series
Participants 175 children and adults with intractable temporal lobe epilepsy at 1 American centre
Interventions Anterior temporal lobectomy with amygdalohippocampectomy with resection of the lateral temporal
cortex and the mesial temporal structures, which included the amygdala, the hippocampus, and the
parahippocampal gyrus
Notes —

Rausch 2003
Methods Not stated - but a longitudinal study, possibly prospective
Participants 44 participants who had a temporal lobe resection from a US centre
Interventions Temporal lobe resection
Notes —

Informed decisions.


Remi 2011
Methods Retrospective case study
Participants 154 German participants with a focal epilepsy, temporal and extratemporal
Interventions Resections were tailored so as to encompass as much epileptogenic tissue as possible and as little elo-
quent cortex as possible
Notes —

Roberti 2007
Methods Retrospective case study
Participants 42 adults and children with non-lesional temporal lobe epilepsy from a single US centre
Interventions All had anteromedial temporal lobectomy
Outcomes Seizure outcome by Engel class at a median of 60 months but minimum of 12 months follow-up
Notes —

Rossi 1994
Participants 138 children and adults with intractable temporal and extratemporal lobe epilepsy at 1 Italian centre
Interventions The surgical procedures utilised were: 17 hemispherectomies (children only), 67 anterior temporal
lobectomies and 54 extratemporal resections
Notes —

Russo 2003
Methods Retrospective case series from 1 Italian centre
Participants 126 participants, adults and children, with age at surgery ranging from 0 to 53 years, with an intractable
epilepsy and malformation of cortical development on histology
Notes —

Informed decisions.


Sagher 2012
Participants 96 patients with medically refractory mesial temporal lobe epilepsy
Outcomes Seizure outcome measured using Engel's classification at 3 months, 1 year, 2 years and 3 years post-
surgery
Notes —

Sakamoto 2009
Methods
Participants
Interventions
Outcomes
Notes

Salanova 1994
Participants 98 children and adults with intractable temporal lobe epilepsy at 1 American centre
Notes —

Sarkis 2012
Participants 63 participants, aged 1.6 to 56 years at time of surgery, who had a multi-lobar resection for a medically
refractory epilepsy
Interventions Multilobar surgical resections, classified based on the lobes involved (frontal, parietal, temporal or oc-
cipital) and then categorised into frontotemporal (FT), temporoparietal (TP), frontoparietal (FP) and
occipital plus (temporoparietooccipital (TPO), parietooccipital (PO), or temporooccipital (TO)). The oc-
cipital plus group represented extended posterior quadrant resections as opposed to the more anterior
resection subsets

Informed decisions.

Sarkis 2012 (Continued)
Outcomes Seizure outcome measured using Engel's classification 1 year post-surgery and annually thereafter
Notes —

Schramm 2011
Methods Randomised controlled trial
Participants 207 participants from 3 German centres with temporal lobe epilepsy
Interventions Randomised either to an intended minimum resection length of 25 mm or 35 mm for the hippocampus
and parahippocampus
Notes —

Seymour 2012
Methods Retrospective case series from a UK centre
Participants 306 adults and children operated on between 1975 and 1995
Interventions Resection of either temporal lobe
Notes A report on mortality, after a longer interval, in a cohort treated by temporal lobe surgery between
1975 and 1995

Sinclair 2003
Methods Retrospective case series from 1 Canadian centre
Participants 77 children with intractable epilepsy
Interventions Extratemporal operations: 8 parietal, 12 frontal, 4 occipital and 10 multi-lobar or hemispherectomy re-
sections. One hypothalamic hamartoma and four callosotomies (reported separately). There were 42
temporal lobectomies (method not defined in any detail)
Outcomes Seizure outcome was assessed using Engel's classification at 1 year postoperatively and annually
thereafter
Notes —

Sindou 2006
Methods Not stated - probably retrospective case series

Informed decisions.

Sindou 2006 (Continued)
Participants 100 people with medically intractable temporo-mesial epilepsy from a French centre
Interventions Tailored temporal lobe resection
Outcomes Seizure outcome was assessed using Engel's classification at least 1 year postoperatively up to 10 years
follow-up
Notes —

Sola 2005
Methods Retrospective case series; multicentre study from Spain
Participants 137 teenagers and adults followed for 2 years postoperatively
Outcomes Seizure outcome by Engel class (1 and 2 combined) determined at 24 months after surgery
Notes —

Spencer 2005
Methods Prospective observational multi-centre American study of seizures, anxiety, depression and quality of
life (QOL) outcomes after resective epilepsy surgery
Participants 339 participants followed for at least 2 years
Interventions Resective surgery: mesial temporal resection or resection in any neocortical region including temporal
lobe
Outcomes Any seizure at 12 months or 1 to 2 years postoperatively
Notes —

Sperling 1992
Participants 51 adults with intractable temporal lobe epilepsy at 1 American centre
Interventions Standard anterior temporal lobectomy. In the non-dominant hemisphere the resection line ran 5.0 cm
to 5.5 cm from the temporal tip and 4.5 cm to 5.0 cm in the dominant hemisphere. The amygdala and
1.5 cm to 2.0 cm of the hippocampus were removed by suction in early and en bloc in later participants
in the series
Notes —

Informed decisions.


Stavrou 2008
Participants 53 children and adults with a cavernoma and epilepsy from single Austrian centre
Interventions 58 microsurgical resections; 3 underwent several operations for multiple cavernomas; 1 participant un-
derwent a second operation for remaining lesion 4 years after initial surgery
Outcomes Seizure outcome by ILAE classification at least 2 years after surgery
Notes —

Suppiah 2009
Participants 176 children and adults from New Zealand followed up for at least 12 months
Notes —

Swartz 1992
Interventions Temporal lobe resection
Outcomes Seizure outcome assessed by seizure freedom Engel class at least 20 months post-surgery
Notes —

Tanriverdi 2010
Participants 256 participants from Canada and France with mesial temporal lobe epilepsy
Interventions Corticohippocampectomy or selective amygdalohippocampectomy
Notes A comparison of surgical approach and IQ and memory outcomes at the 1-year follow-up in people
with medically refractory mesial temporal lobe epilepsy (MTLE) due to hippocampal sclerosis

Informed decisions.


Tatum 2008
Methods Retrospective case series from 2 US centres
Participants 39 adults with an intractable localisation-related epilepsy and a normal MR scan
Outcomes Participants were classified as seizure-free or not seizure-free at follow-up, which was at least 12
months after surgery
Notes —

Terra-Bustamante 2005a
Methods Prospective study from a Brazilian centre
Participants 107 patients, 18 years and under with a medically intractable epilepsy
Outcomes Seizure outcome was classified according to Engel's classification scheme and assessed at least 12
months postoperatively
Notes —

Terra-Bustamante 2005b
Methods Prospective study from a Brazilian centre
Participants 35 children, 18 years and under with a medically intractable temporal lobe epilepsy
Interventions All had surgery on the temporal lobe
Notes —

Tezer 2008
Methods Retrospective case series from a Turkish centre
Participants 109 adults with mesial temporal lobe epilepsy and hippocampal sclerosis
Interventions Anterior temporal lobectomy in all
Outcomes Seizure outcome assessed using Engel's classification at least 12 months after surgery
Notes —

Informed decisions.


Theodore 2012
Participants 41 adults from 1 American centre
Interventions All participants underwent anterior temporal lobectomy, tailored to individual pre-resection evalua-
tions and intraoperative electrocorticography. All resections included the temporal tip, a minimum of
1 cm of the anterior part of the superior temporal gyrus, and between 3 cm and 5 cm of the middle and
inferior temporal gyri. Resection was extended to involve epileptogenic frontal regions identified on
subdural electrode recording in 2
Notes The objective of this study was to compare 5-hydroxytryptamine receptor 1A (5-HT1A) PET with cere-
bral metabolic rate of glucose (CMRglc) PET for temporal lobectomy planning

Tigaran 2003
Participants 65 adults who had surgery for intractable partial epilepsy
Interventions All had frontal lobe cortical resections
Notes —

Tripathi 2008
Methods Retrospective cases series from an Indian centre
Participants 57 children and adults (61% under 18 years) with intractable epilepsy secondary to cortical dysplasia
Outcomes Seizure outcome measured at least 12 months postoperatively classified using Engel's scale
Notes —

Trottier 2008
Participants 105 French adults (2 children) from 1 centre
Interventions Lesionectomies defined by MRI/EEG concordance or by SEEG

Informed decisions.

Trottier 2008 (Continued)
Outcomes Seizure outcome assessed using Engel's classification at follow-up at 12 months after surgery and an-
nually thereafter
Notes —

Urbach 2007
Participants 42 adults and children with drug-resistant parietal and occipital lobe epilepsies
Outcomes Seizure outcome at 12 months was determined using Engel's classification
Notes —

Ure 2009
Participants 77 participants with bitemporal lobe epilepsy aged 14 to 53 years from 1 Canadian centre
Interventions Target temporal lobectomies
Outcomes Seizure outcome was assessed using Engel's classification at 1 year
Notes A study of the usefulness of intracranial electrical stimulation in identifying the temporal lobe to be tar-
geted for resection in bitemporal lobe epilepsy

Velasco 2011
Methods Prospective case series from a Brazilian centre
Participants 163 patients, over 18 years old, with refractory mesial temporal lobe epilepsy and hippocampal sclero-
sis
Outcomes Seizure outcome measured at least 14 months after surgery using Engel's classification
Notes —

Walz 2003
Methods Not stated whether prospective or retrospective
Participants 100 adults with mesial temporal lobe epilepsy related to hippocampal sclerosis from a Brazilian centre

Informed decisions.

Walz 2003 (Continued)
Interventions Anterior or mesial temporal lobectomy
Notes —

Weinand 1992
Participants 89 participants (age not stated) with intractable temporal lobe epilepsy at 1 American centre
Interventions For participants with medial temporal lobe onset a standardised 4.5 cm lateral resection with the pos-
terior resection line of the parahippocampal gyrus and hippocampus extending to at least the level of
the cerebral peduncle; for lateral onset a more extensive and tailored ECoG-guided resection; for re-
gional temporal lobe onset more extensive removal of the parahippocampal gyrus, hippocampus and
lateral cortex
Notes —

Wellmer 2012
Participants 197 participants, aged between 0 and 70 years, with a pharmacoresistant epilepsy who had invasive
monitoring prior to surgery
Interventions Not stated
Outcomes Seizure outcome measured at 3, 6, 12 and 24 months after surgery using Engel's classification
Notes —

Widdess-Walsh 2007
Participants 48 participants (30 were over 16 years) undergoing surgery for focal cortical dysplasia guided by subd-
ual electrode recordings
Outcomes Measurement of seizure freedom at follow-up at least 12 months from time of surgery
Notes —


Informed decisions.

Wiebe 2001
Methods Randomised controlled trial conducted in 3 centres in 1 Canadian city
Participants Participants had to be at least 16 years old and to have had seizures with strong temporal lobe semiolo-
gy for more than 1 year
Interventions Randomisation (40 in each group) to a resection of a maximum of 6.0 cm to 6.5 cm of the anterior later-
al nondominant temporal lobe or 4.0 cm to 4.5 cm of the dominant temporal lobe. The mesial resection
included the amygdala and, at a minimum, the anterior 1.0 cm to 3.0 cm of the hippocampus (most
commonly, 4.0 cm) OR continuing on antiepileptic drugs
Outcomes The primary outcome was freedom from seizures that impair awareness of self and surroundings at 1
year
Notes —

Wieshmann 2008
Methods Retrospective case series from a UK centre
Participants 76 adults with refractory temporal lobe epilepsy
Interventions Anterior temporal lobe resections
Outcomes Seizure outcome assessed 2 years post-surgery using Engel's classification
Notes —

Wray 2012
Participants 52 children from one American centre; all participants had the primary motor or somatosensory cortex
localised via 2 or more of the following tests: SSEP, fMRI or high gamma electrocorticography (hgECoG)
Interventions Resection of epileptogenic zone: temporal and extratemporal
Notes —

Wyler 1995
Interventions The same procedure for either hemisphere and all participants: the anterior lateral 4.5 cm of tempo-
ral neocortex was removed, including the superior through inferior temporal gyri, leaving hippocam-
pus, parahippocampus and fusiform gyri. Fusiform gyrus was then dissected piecemeal, leaving only
hippocampal and parahippocampal gyri. The hippocampus was removed en bloc to either the anterior

Informed decisions.

Wyler 1995 (Continued)
margin of the cerebral peduncle (partial hippocampectomy (P)) or to the level of the colliculi (total hip-
pocampectomy (T))
Notes —

Wyllie 1998
Methods Case series (whether prospective or retrospective not stated)
Participants 136 children with intractable extratemporal or temporal lobe epilepsy at 1 American centre
Interventions Participants with hippocampal sclerosis had anteromesial temporal resection. For other temporal le-
sions a resection of the lesion, surrounding cortex
and usually also the mesial temporal structures was performed. 46% of extratemporal resections were
frontal and the rest were parietal, occipital, perirolandic or multilobar (frontal and temporal, temporal
and occipital, or temporo-parietal-occipital). The functional hemispherectomies were performed as de-
scribed by Rasmussen with resection of central regions and hemispheric disconnection by transection
of white matter tracts and corpus callosotomy
Notes —

Yang 2011
Participants 99 children and adults followed for at least 1 year from 1 Chinese centre
Interventions Lesion resection defined by clinical, neuroimaging and electrophysiological results
Outcomes Seizure outcome assessed at least 12 months post-surgery using modified ILAE classification
Notes —

Yeon 2009
Methods Retrospective case series from a Korean centre
Participants 60 adults with an infratentorial cavernous haemangioma and seizures
Notes —

Informed decisions.


Yu 2009
Methods Retrospective case series from a Chinese centre
Participants 43 adults and children with a posterior cortex epilepsy
Notes —

Yu 2012a
Participants 100 adults with resective epilepsy surgery between 2001 and 2009 in 1 Chinese centre
Interventions Classic anterior temporal lobotomies were performed for temporal lobe epilepsy. Tailored resection
was completed for patients with extratemporal lobe epilepsy. Resective micro-surgeries were con-
ducted with the guidance of presurgical localisation results and aimed to remove epileptogenic zones.
Anatomic or functional hemispherectomy was performed on certain patients with hemispheric lesions.
62 with temporal lobe epilepsy had anterior temporal lobectomies, 37 had tailored focal or lobar resec-
tions with extra temporal lobe epilepsy (6 with tumours). 1 had anatomic hemispherectomy
Outcomes Seizure outcome assessed at least 12 months post-surgery using Engel classification
Notes Yu 2012a and Yu 2012b were treated as 2 studies in a single publication. One quality assessment is
made for the single publication

Yu 2012b
Participants 222 children with resective epilepsy surgery between 2001 and 2009 in 1 Chinese centre
Interventions Classic anterior temporal lobotomies were performed for temporal lobe epilepsy. Tailored resection
was completed for patients with extratemporal lobe epilepsy. Resective micro-surgeries were con-
ducted with the guidance of presurgical localisation results and aimed to remove epileptogenic zones.
Anatomic or functional hemispherectomy was performed on certain patients with hemispheric lesions.
62 with temporal lobe epilepsy had anterior temporal lobectomies, 37 had tailored focal or lobar resec-
tions with extra temporal lobe epilepsy (6 with tumours). 1 had anatomic hemispherectomy
Outcomes Seizure outcome assessed at least 12 months post-surgery using Engel classification
Notes Yu 2012a and Yu 2012b were treated as 2 studies in a single publication. One quality assessment is
made for the single publication

Zangaladze 2008
Methods Both a retrospective and prospective case series from a US centre

Informed decisions.

Zangaladze 2008 (Continued)
Participants 99 participants 12 years and older with a localisation-related epilepsy who had had intracranial EEG
recordings prior to surgery
Interventions Temporal and extratemporal lobe resections
Outcomes Seizure freedom at follow-up, which was no less than 2 years after surgery
Notes —

Zentner 1995
Participants 178 children and adults with intractable temporal lobe epilepsy at a single German centre
Interventions Procedures performed: anterior temporal lobectomy (standard or "keyhole") with hippocampectomy,
anterior temporal lobectomy without hippocampectomy extended lesionectomy with hippocampecto-
my, extended lesionectomy without hippocampectomy and selective amygdalohippocampectomy
Notes —

Zentner 1996
Participants 60 children and adults with intractable extratemporal lobe epilepsy at a single German centre
Interventions The following surgical procedures were performed: frontal lobectomy (n = 16), frontal topectomy (n =
24), parietal topectomy (n = 7) and occipital topectomy
Notes —
ATL: anterior temporal lobectomy

BMI: body mass index
ECoG: electrocorticography
EEG: electroencephalography
EZ: Epileptogenic Zone
FDG: fluorodeoxyglucose
fMRI: functional magnetic resonance imaging
HS: Hippocampal Sclerosis
ICEEG: intracranial electroencephalography
ILAE: International League Against Epilepsy
MR: magnetic resonance
MRI: magnetic resonance imaging
MTLE: mesial temporal lobe epilepsy
PET: positron emission tomography
PO: Postoperative
SD: standard deviation
SEEG: stereoelectroencephalography
Informed decisions.

SPECT: single photon emission computed tomography

SSEP: somatosensory evoked potential
TLE: temporal lobe epilepsy

Characteristics of excluded studies [ordered by study ID]

Study Reason for exclusion
Acar 2008 Only 17 participants followed up for more than 12 months
Alemany-Rosales 2011 No seizure outcome data reported
Alpherts 2008 Study has only 6 months follow-up
Andersson-Roswall 2010 No seizure outcome data reported
Asadi-Pooya 2008 A study of postoperative antiepileptic drug treatment rather than seizure outcome of surgery
Bauer 2007 The study has fewer than 30 participants
Baxendale 2005 No seizure outcome data reported
Bell 2010 No seizure outcome data reported
Binder 2009 Study has minimum 4-month follow-up; unknown number followed up for 1 year
Boesebeck 2002 Subset of the participants reported in Boesebeck 2007
Boshuisen 2012 Only those with 12 postoperative months seizure-free were included in the study
Bourgeois 2007 The study has fewer than 30 participants
Buckingham 2010 Only participants with postoperative seizures were included in the study
Busch 2011 Study has only 5 to 7 months postoperative follow-up
Caicoya 2007 The study has fewer than 30 participants
Carne 2004 No seizure outcome data reported
Cascino 1996 Reports the same participants as Cascino 1995, more relevant information in Cascino 1995
Chang 2007 An unknown proportion were followed up for less than 1 year in the study
Choi 2004b No follow-up period defined for the study
Cohen-Gadol 2003 An unknown proportion were followed up for less than 1 year in the study
Colonnelli 2012 No follow-up period defined for the study
Coutin-Churchman 2012 No follow-up period defined for the study
Cukiert 2009 The study has fewer than 30 participants
D'Angelo 2006 An unknown proportion were followed up for less than 1 year in the study
D'Argenzio 2011 No seizure outcome data reported

Informed decisions.

da Costa-Neves 2012 Study follow-up is 6 months
Danielsson 2009 The study has fewer than 30 participants
Datta 2009 The study has fewer than 30 participants
Dulay 2006 The study has less than a 12-month follow-up period
Dulay 2009 Postoperative follow-up period for most participants was less than 12 months
Elsharkawy 2008b Reports on a subset of the participants from Elsharkawy 2009a
Elsharkawy 2009b Questionnaire follow-up study with no seizure outcome data
Elsharkawy 2011b The study only recruits children not seizure-free 6 months postoperatively
Engel Jr 2012 The study has fewer than 30 participants
Fauser 2008 No seizure outcome data reported
Ferrari-Marinho 2012 No postoperative follow-up period defined for the study
Ferroli 2006 An unknown proportion of participants were followed up for fewer than 12 months
Freitag 2005 Study follow-up defined as 6 to 12 months
Ghacibeh 2009 Study follow-up period is not defined
Griffin 2007 No seizure outcome data reported
Grunert 2003 No duration of outcome and not surgery for epilepsy (surgery for tumours)
Haegelen 2013 Fewer than 30 participants in the study have a 12-month follow-up period
Harvey 2008 Follow-up period of the study was 6 months
Hellwig 2012 Follow-up period of the study was less than 12 months
Helmstaedter 2004 Study with a 3-month follow-up period
Helmstaedter 2011 Reports a subset of the participants from Schramm 2011 (results from a single centre of Schramm
2011)
Hervas-Navidad 2002 Some participants followed up for only 6 months and no seizure outcome data reported for some
participants
Hildebrandt 2005 No seizure outcome data reported
Hu 2012 An unknown proportion of participants were followed up for less than 12 months
Jehi 2010 The study only included those who experienced postoperative seizures
Junna 2013 Seizure outcome is not measured at 1 year for all participants and analyses cannot be separated
Kuzniecky 1996 Reports on the same participants as Kuzniecky 1993, more relevant information in Kuzniecky 1993

Informed decisions.

Lach 2010 Case control design (participants selected based on seizure outcome) is inappropriate for this re-
view
Lachhwani 2003 Commentary on Wiebe 2001
Lee 2010 The study has fewer than 30 participants
Limbrick 2009 An unknown proportion of participants were followed up for less than 12 months
Lodenkemper 2007 Seizure outcome data are reported for fewer than 30 participants
Lutz 2004 No seizure outcome data reported
Malla 1998 Reports the same participants as Cascino 1995, more relevant information in Cascino 1995
McClelland 2007 No seizure outcome data reported
McClelland 2011 No seizure outcome data reported
Mikati 2004 The study has fewer than 30 participants
Mohammed 2012 Fewer than 90% had MRI scans
Moien-Afshari 2009 The study has fewer than 30 participants
Negishi 2011 The study has fewer than 30 participants
Nikase 2007 The study has fewer than 30 participants
Oertel 2004 No seizure outcome data reported
Ogiwara 2010 An unknown proportion of participants were followed up for less than 12 months
Park 2010 Seizure outcome is measured following antiepileptic drug reduction rather than outcome after
surgery
Placantonakis 2010 The study has fewer than 30 participants
Rocamora 2009 The study has fewer than 30 participants
Roth 2011 An unknown proportion of participants were followed up for less than 12 months
Sakuta 2005 The study has fewer than 30 participants
Schwartz 2006 Data are reported only for those participants who were seizure-free at 12 months
Smyth 2007 An unknown proportion of participants were followed up for less than 12 months
Soeder 2009 The study has fewer than 30 participants
Stavem 2005 No seizure outcome data reported
Stavem 2008 No seizure outcome data reported
Stefan 2004 An unknown proportion of participants were followed up for less than 12 months

Informed decisions.

Stefan 2008 Case-control design (participants selected based on seizure outcome) is inappropriate for this re-
view
Teutonico 2008 The study has fewer than 30 participants
Upchurch 2010 The study has fewer than 30 participants
Vachrajani 2012 An unknown proportion of participants were followed up for less than 12 months
Vadera 2012 An unknown proportion of participants were followed up for less than 12 months
Vadlamudi 2004 Only seizure-free participants are included in the study
Weinand 2001 Reports on the same participants as Weinand 1992, more relevant information given in Weinand
1992
Wetjen 2006 No seizure outcome data reported
Wetjen 2009 The study has fewer than 30 participants
Wieser 2003a Fewer than 90% had a MRI scan
Wieser 2003b Fewer than 90% had a MRI scan
Yasuda 2010a No seizure outcome data reported
Yasuda 2010b The study has fewer than 30 participants
Zupanc 2010 An unknown proportion of participants were followed up for less than 12 months

DATA AND ANALYSES

Comparison 1. Surgery versus medical treatment (randomised evidence)
Outcome or subgroup title No. of No. of Statistical method Effect size

studies partici-
pants
1 Proportion free from seizure impairing awareness at 1 80 Risk Ratio (M-H, Fixed, 95% 7.67 [2.50,
one year CI) 23.51]
2 Proportion free from all seizures (including auras) at 1 80 Risk Ratio (M-H, Fixed, 95% 15.0 [2.08,
one year CI) 108.23]


Informed decisions.

Analysis 1.1. Comparison 1 Surgery versus medical treatment (randomised

evidence), Outcome 1 Proportion free from seizure impairing awareness at one year.
Study or subgroup Surgery Medical Risk Ratio Weight Risk Ratio
treatment
n/N n/N M-H, Fixed, 95% CI M-H, Fixed, 95% CI
Wiebe 2001 23/40 3/40 100% 7.67[2.5,23.51]

Total (95% CI) 40 40 100% 7.67[2.5,23.51]
Total events: 23 (Surgery), 3 (Medical treatment)
Heterogeneity: Not applicable
Test for overall effect: Z=3.56(P=0)
Medical treatment 0.01 0.1 1 10 100 Surgery

Analysis 1.2. Comparison 1 Surgery versus medical treatment (randomised
evidence), Outcome 2 Proportion free from all seizures (including auras) at one year.
Study or subgroup Surgery Medical Risk Ratio Weight Risk Ratio
treatment
Wiebe 2001 15/40 1/40 100% 15[2.08,108.23]

Total (95% CI) 40 40 100% 15[2.08,108.23]
Total events: 15 (Surgery), 1 (Medical treatment)
Test for overall effect: Z=2.69(P=0.01)
Medical treatment 0.01 0.1 1 10 100 Surgery

Comparison 2. Comparison of surgical techniques (randomised evidence)

studies partici-
pants
1 Proportion free from seizures (Engel Class 1) at 2 1 60 Risk Ratio (M-H, Fixed, 95% CI) 1.22 [0.85, 1.76]
years
2 Proportion free from seizures (Engel Class 1) at 1 1 207 Risk Ratio (M-H, Fixed, 95% CI) 1.02 [0.86, 1.20]
year
3 Proportion free from all seizures (including auras) 1 70 Risk Ratio (M-H, Fixed, 95% CI) 1.82 [1.12, 2.93]
at 1 year


Informed decisions.

Analysis 2.1. Comparison 2 Comparison of surgical techniques (randomised

evidence), Outcome 1 Proportion free from seizures (Engel Class 1) at 2 years.
Study or subgroup ATL with cor- Anterior tem- Risk Ratio Weight Risk Ratio
pus calloso- poral lobec-
tomy (aCCT) tomy (ATL)
Liang 2010 22/30 18/30 100% 1.22[0.85,1.76]

Total (95% CI) 30 30 100% 1.22[0.85,1.76]
Total events: 22 (ATL with corpus callosotomy (aCCT)), 18 (Anterior tempo-
ral lobectomy (ATL))
Favours ATL 0.01 0.1 1 10 100 Favours aCCT

evidence), Outcome 2 Proportion free from seizures (Engel Class 1) at 1 year.
Study or subgroup 2.5 cm re- 3.5 cm re- Risk Ratio Weight Risk Ratio
section section
Schramm 2011 77/104 75/103 100% 1.02[0.86,1.2]

Total (95% CI) 104 103 100% 1.02[0.86,1.2]
Total events: 77 (2.5 cm resection), 75 (3.5 cm resection)
Favours 3.5 cm resection 0.01 0.1 1 10 100 Favours 2.5 cm resection

evidence), Outcome 3 Proportion free from all seizures (including auras) at 1 year.
Study or subgroup Total hip- Partial hip- Risk Ratio Weight Risk Ratio
pocampectomy pocampectomy
Wyler 1995 25/36 13/34 100% 1.82[1.12,2.93]

Total (95% CI) 36 34 100% 1.82[1.12,2.93]
Total events: 25 (Total hippocampectomy), 13 (Partial hippocampectomy)
Heterogeneity: Tau2=0; Chi2=0, df=0(P<0.0001); I2=100%
Favours partial 0.01 0.1 1 10 100 Favours total


Informed decisions.

Comparison 3. Surgery for epilepsy (randomised and non-randomised evidence)

studies partici-
pants
1 Proportion with a good outcome of 173 Std. Mean Difference (Fixed, 95% CI) Totals not selected
surgery
1.1 > 1 year seizure-free 42 Std. Mean Difference (Fixed, 95% CI) 0.0 [0.0, 0.0]
1.2 Engel Class Scale 114 Std. Mean Difference (Fixed, 95% CI) 0.0 [0.0, 0.0]
1.3 Other scale 17 Std. Mean Difference (Fixed, 95% CI) 0.0 [0.0, 0.0]

Analysis 3.1. Comparison 3 Surgery for epilepsy (randomised and non-
randomised evidence), Outcome 1 Proportion with a good outcome of surgery.
Study or subgroup Poor outcome Good outcome Std. Mean Std. Mean Difference Std. Mean Difference
Difference
N N (SE) IV, Fixed, 95% CI IV, Fixed, 95% CI
3.1.1 > 1 year seizure-free
Adelson 1992 10 23 69.7 (8) 69.7[54.02,85.38]
Althausen 2013 16 45 73.8 (5.632) 73.77[62.73,84.81]
Awad 1991 20 27 57.5 (7.212) 57.45[43.31,71.59]
Bautista 2003 15 28 65.1 (7.268) 65.12[50.87,79.37]
Chee 1993 10 28 73.7 (7.143) 73.68[59.68,87.68]
Dagar 2011 23 89 79.5 (3.817) 79.46[71.98,86.94]
Duchowny 1998 15 16 51.6 (8.976) 51.61[34.02,69.2]
Erba 1992 9 37 80.4 (5.849) 80.43[68.97,91.89]
Fujiwara 2012 21 23 52.3 (7.53) 52.27[37.51,67.03]
Garcia 1991 20 35 63.6 (6.486) 63.64[50.93,76.35]
Garcia 1994 15 36 70.6 (6.38) 70.59[58.08,83.1]
Gilliam 1997a 25 53 68 (5.284) 67.95[57.59,78.31]
Gilliam 1997b 11 22 66.7 (8.206) 66.67[50.59,82.75]
Greiner 2011 12 42 77.8 (5.658) 77.78[66.69,88.87]
Hallbook 2010 22 88 80 (3.814) 80[72.52,87.48]
Hemb 2010 102 223 68.6 (2.574) 68.62[63.57,73.67]
Holmes 1997 22 22 50 (7.538) 50[35.23,64.77]
Jack 1992 16 34 68 (6.597) 68[55.07,80.93]
Jeha 2006 140 231 62.3 (2.517) 62.26[57.33,67.19]
Jehi 2012 147 165 52.9 (2.826) 52.88[47.34,58.42]
Jennum 1993 22 42 65.6 (5.937) 65.63[53.99,77.27]
Kim 2009 72 94 56.6 (3.847) 56.63[49.09,64.17]
Kim 2010b 26 14 35 (7.542) 35[20.22,49.78]
Lei 2008 16 180 91.8 (1.956) 91.84[88.01,95.67]
Lopez-Gonzalez 2012 21 65 75.6 (4.633) 75.58[66.5,84.66]
Morino 2009 7 55 88.7 (4.019) 88.71[80.83,96.59]
O'Brien 1996 10 36 78.3 (6.082) 78.26[66.34,90.18]
Oertel 2005 15 20 57.1 (8.365) 57.14[40.75,73.53]
Paolicchi 2000 31 44 58.7 (5.686) 58.67[47.53,69.81]
Rausch 2003 21 21 50 (7.715) 50[34.88,65.12]
Poor outcome -100 -50 0 50 100 Good outcome

Informed decisions.

Difference
Spencer 2005 91 264 74.4 (2.317) 74.37[69.83,78.91]
Swartz 1992 7 27 79.4 (6.934) 79.41[65.82,93]
Tatum 2008 17 22 56.4 (7.94) 56.41[40.85,71.97]
Theodore 2012 15 26 63.4 (7.522) 63.41[48.67,78.15]
Walz 2003 13 85 86.7 (3.426) 86.73[80.01,93.45]
Weinand 1992 32 57 64 (5.087) 64.04[54.07,74.01]
Widdess-Walsh 2007 26 22 45.8 (7.192) 45.83[31.73,59.93]
Wiebe 2001 13 23 63.9 (8.005) 63.89[48.2,79.58]
Wyler 1995 32 38 54.3 (5.954) 54.29[42.62,65.96]
Yu 2012a 39 61 61 (4.878) 61[51.44,70.56]
Yu 2012b 74 148 66.7 (3.164) 66.67[60.47,72.87]
Zangaladze 2008 33 66 66.7 (4.738) 66.67[57.38,75.96]

3.1.2 Engel Class Scale
Aaberg 2012 24 30 55.6 (6.762) 55.56[42.31,68.81]
Adam 1996 4 26 86.7 (6.206) 86.67[74.51,98.83]
Alfstad 2011 23 25 52.1 (7.211) 52.08[37.95,66.21]
Arruda 1996 21 53 71.6 (5.241) 71.62[61.35,81.89]
Babini 2013 4 26 86.7 (6.206) 86.67[74.51,98.83]
Battaglia 2006 13 32 71.1 (6.757) 71.11[57.87,84.35]
Baumann 2007 50 118 70.2 (3.527) 70.24[63.33,77.15]
Bell 2009 16 24 60 (7.746) 60[44.82,75.18]
Benifla 2006 28 78 73.6 (4.282) 73.58[65.19,81.97]
Berkovic 1995 61 74 54.8 (4.283) 54.81[46.41,63.21]
Blount 2004 18 12 40 (8.944) 40[22.47,57.53]
Blume 2004 44 26 37.1 (5.775) 37.14[25.82,48.46]
Brainer-Lima 1996 3 29 90.6 (5.153) 90.63[80.53,100.73]
Britton 1994 17 34 66.7 (6.601) 66.67[53.73,79.61]
Caraballo 2011 12 33 73.3 (6.592) 73.33[60.41,86.25]
Cascino 1995 52 113 68.5 (3.617) 68.48[61.39,75.57]
Chabardes 2005 7 47 87 (4.571) 87.04[78.08,96]
Chang 2009 12 45 79 (5.4) 78.95[68.37,89.53]
Chkhenkeli 2007 55 12 17.9 (4.684) 17.91[8.73,27.09]
Choi 2004a 8 27 77.1 (7.098) 77.14[63.23,91.05]
Chung 2005 70 58 45.3 (4.4) 45.31[36.69,53.93]
Cossu 2005 72 93 56.4 (3.861) 56.36[48.79,63.93]
Cossu 2008 36 77 68.1 (4.383) 68.14[59.55,76.73]
Costello 2009 10 32 76.2 (6.572) 76.19[63.31,89.07]
Cukiert 2002 11 89 89 (3.129) 89[82.87,95.13]
Dalmagro 2005 15 28 65.1 (7.268) 65.12[50.87,79.37]
Delbeke 1996 15 23 60.5 (7.929) 60.53[44.99,76.07]
Dellabadia 2002 15 20 57.1 (8.365) 57.14[40.75,73.53]
Devlin 2003 16 17 51.5 (8.7) 51.52[34.47,68.57]
Donadio 2011 24 60 71.4 (4.929) 71.43[61.77,81.09]
Dorward 2011 15 18 54.6 (8.668) 54.55[37.56,71.54]
Dunkley 2011 22 20 47.6 (7.706) 47.62[32.52,62.72]
Dunlea 2010 80 119 59.8 (3.476) 59.8[52.99,66.61]
Elsharkawy 2008a 97 121 55.5 (3.366) 55.5[48.9,62.1]
Elsharkawy 2009a 119 311 72.3 (2.158) 72.33[68.1,76.56]
Elsharkawy 2011a 9 38 80.9 (5.739) 80.85[69.6,92.1]
Erickson 2005 25 46 64.8 (5.668) 64.79[53.68,75.9]

Informed decisions.

Difference
Fauser 2004 24 35 59.3 (6.395) 59.32[46.79,71.85]
Gelinas 2011 15 52 77.6 (5.093) 77.61[67.63,87.59]
Georgakoulias 2008 8 42 84 (5.185) 84[73.84,94.16]
Grivas 2006 15 37 71.2 (6.283) 71.15[58.84,83.46]
Hader 2004 18 21 53.9 (7.983) 53.85[38.2,69.5]
Hajek 2009 10 25 71.4 (7.636) 71.43[56.46,86.4]
Hamiwka 2005 21 17 44.7 (8.066) 44.74[28.93,60.55]
Hartley 2002 15 20 57.1 (8.365) 57.14[40.75,73.53]
Hartzfield 2008 32 24 42.9 (6.613) 42.86[29.9,55.82]
Janszky 2003a 24 60 71.4 (4.929) 71.43[61.77,81.09]
Janszky 2003b 24 123 83.7 (3.048) 83.67[77.7,89.64]
Jaramillo-Betancur 2009 24 43 64.2 (5.858) 64.18[52.7,75.66]
Jayakar 2008 57 44 43.6 (4.934) 43.56[33.89,53.23]
Jayalakshmi 2011 28 50 64.1 (5.432) 64.1[53.45,74.75]
Kan 2008 15 43 74.1 (5.75) 74.14[62.87,85.41]
Kang 2009 50 194 79.5 (2.584) 79.51[74.45,84.57]
Kanner 2009 14 86 86 (3.47) 86[79.2,92.8]
Kilpatrick 1997 11 39 78 (5.858) 78[66.52,89.48]
Kim 2010a 102 75 42.4 (3.714) 42.37[35.09,49.65]
Kloss 2002 34 34 50 (6.063) 50[38.12,61.88]
Knowlton 2008 25 37 59.7 (6.23) 59.68[47.47,71.89]
Lee 2008 33 38 53.5 (5.919) 53.52[41.92,65.12]
Lee 2010a 16 36 69.2 (6.4) 69.23[56.69,81.77]
Lee 2011 3 37 92.5 (4.165) 92.5[84.34,100.66]
Li 1997 12 39 76.5 (5.94) 76.47[64.83,88.11]
Li 1999 22 16 42.1 (8.009) 42.11[26.41,57.81]
Liang 2010 20 40 66.7 (6.086) 66.67[54.74,78.6]
Liang 2012 33 173 84 (2.556) 83.98[78.97,88.99]
Liava 2012 13 39 75 (6.005) 75[63.23,86.77]
Lorenzo 1995 31 17 35.4 (6.903) 35.42[21.89,48.95]
Madhavan 2007 33 37 52.9 (5.966) 52.86[41.17,64.55]
Mani 2006 36 86 70.5 (4.129) 70.49[62.4,78.58]
Mihara 2004 88 269 75.4 (2.281) 75.35[70.88,79.82]
Miserocchi 2013 10 58 85.3 (4.295) 85.29[76.87,93.71]
Morris 1998 10 26 72.2 (7.465) 72.22[57.59,86.85]
O'Brien 2000 22 14 38.9 (8.125) 38.89[22.97,54.81]
Paglioli 2006 17 143 89.4 (2.436) 89.38[84.61,94.15]
Park 2002 40 108 73 (3.65) 72.97[65.82,80.12]
Park 2006 10 20 66.7 (8.607) 66.67[49.8,83.54]
Perego 2009 8 29 78.4 (6.768) 78.38[65.12,91.64]
Perry 2010 49 34 41 (5.398) 40.96[30.38,51.54]
Phi 2009 7 80 92 (2.916) 91.95[86.23,97.67]
Pinheiro-Martins 2012 37 33 47.1 (5.966) 47.14[35.45,58.83]
Prevedello 2000 31 53 63.1 (5.265) 63.1[52.78,73.42]
Raabe 2012 19 57 75 (4.967) 75[65.26,84.74]
Radhakrishnan 1998 41 134 76.6 (3.202) 76.57[70.29,82.85]
Remi 2011 16 138 89.6 (2.459) 89.61[84.79,94.43]
Roberti 2007 15 27 64.3 (7.394) 64.29[49.8,78.78]
Russo 2003 37 64 63.4 (4.794) 63.37[53.97,72.77]
Sagher 2012 11 85 88.5 (3.251) 88.54[82.17,94.91]
Salanova 1994 36 53 59.6 (5.202) 59.55[49.35,69.75]

Informed decisions.

Difference
Sarkis 2012 24 38 61.3 (6.186) 61.29[49.17,73.41]
Schramm 2011 55 152 73.4 (3.07) 73.43[67.41,79.45]
Seymour 2012 138 153 52.6 (2.927) 52.58[46.84,58.32]
Sinclair 2003 20 57 74 (4.997) 74.03[64.24,83.82]
Sindou 2006 15 85 85 (3.571) 85[78,92]
Sola 2005 63 74 54 (4.258) 54.01[45.66,62.36]
Sperling 1992 10 41 80.4 (5.56) 80.39[69.49,91.29]
Stavrou 2008 8 45 84.9 (4.917) 84.91[75.27,94.55]
Suppiah 2009 80 94 54 (3.778) 54.02[46.61,61.43]
Tanriverdi 2010 96 160 62.5 (3.026) 62.5[56.57,68.43]
Terra-Bustamante 2005a 45 62 57.9 (4.772) 57.94[48.59,67.29]
Terra-Bustamante 2005b 8 27 77.1 (7.098) 77.14[63.23,91.05]
Tezer 2008 19 90 82.6 (3.634) 82.57[75.45,89.69]
Tigaran 2003 33 32 49.2 (6.201) 49.23[37.08,61.38]
Tripathi 2008 26 29 52.7 (6.732) 52.73[39.54,65.92]
Trottier 2008 23 73 76 (4.356) 76.04[67.5,84.58]
Urbach 2007 17 25 59.5 (7.574) 59.52[44.68,74.36]
Velasco 2011 70 93 57.1 (3.877) 57.06[49.46,64.66]
Wellmer 2012 59 106 64.2 (3.731) 64.24[56.93,71.55]
Wieshmann 2008 30 46 60.5 (5.607) 60.53[49.54,71.52]
Wray 2012 16 36 69.2 (6.4) 69.23[56.69,81.77]
Wyllie 1998 44 92 67.7 (4.012) 67.65[59.79,75.51]
Yeon 2009 10 50 83.3 (4.811) 83.33[73.9,92.76]
Yu 2009 17 26 60.5 (7.456) 60.47[45.86,75.08]
Zentner 1995 64 103 61.7 (3.762) 61.68[54.31,69.05]
Zentner 1996 26 30 53.6 (6.664) 53.57[40.51,66.63]

3.1.3 Other scale
Boesebeck 2007 48 33 66.7 (4.738) 66.67[57.38,75.96]
Boshuisen 2010 10 33 76.7 (6.443) 76.74[64.11,89.37]
de Tisi 2011 370 245 39.8 (1.974) 39.84[35.97,43.71]
Engman 2004 19 35 64.8 (6.499) 64.81[52.07,77.55]
Goldstein 1996 18 15 45.5 (8.668) 45.45[28.46,62.44]
Gyimesi 2007 26 74 74 (4.386) 74[65.4,82.6]
Holmes 2000 72 54 42.9 (4.409) 42.86[34.22,51.5]
Jeong 1999 15 78 83.9 (3.814) 83.87[76.39,91.35]
Kral 2007 13 27 67.5 (7.406) 67.5[52.99,82.01]
Lackmayer 2013 13 32 71.1 (6.757) 71.11[57.87,84.35]
Lee 2006 14 37 72.6 (6.249) 72.55[60.3,84.8]
Mathern 1999 120 78 39.4 (3.472) 39.39[32.58,46.2]
McIntosh 2012 70 11 13.6 (3.806) 13.58[6.12,21.04]
Phi 2010 21 20 48.8 (7.806) 48.78[33.48,64.08]
Rossi 1994 52 86 62.3 (4.125) 62.32[54.23,70.41]
Sakamoto 2009 17 19 52.8 (8.32) 52.78[36.47,69.09]
Yang 2011 72 27 27.3 (4.476) 27.27[18.5,36.04]


Informed decisions.

Comparison 4. Prognostic factors of a good outcome of surgery (randomised and non-randomised evidence)

studies partici-
pants
1 Good outcome by MRI results 42 3956 Risk Ratio (M-H, Fixed, 95% CI) 0.78 [0.73, 0.82]
1.1 > 1 year seizure-free 17 1691 Risk Ratio (M-H, Fixed, 95% CI) 0.82 [0.74, 0.90]
1.2 Engel Class Scale 21 2054 Risk Ratio (M-H, Fixed, 95% CI) 0.74 [0.69, 0.80]
1.3 Other scale 4 211 Risk Ratio (M-H, Fixed, 95% CI) 0.84 [0.53, 1.32]
2 Good outcome by use of intracranial 21 1547 Risk Ratio (M-H, Fixed, 95% CI) 0.85 [0.78, 0.93]
monitoring (IM)
2.3 Other Scale 1 50 Risk Ratio (M-H, Fixed, 95% CI) 0.53 [0.28, 0.98]
3 Good outcome by presence of 46 4430 Risk Ratio (M-H, Fixed, 95% CI) 1.17 [1.12, 1.23]
mesial temporal sclerosis (MTS)
4 Good outcome by Concordance of 23 1778 Risk Ratio (M-H, Fixed, 95% CI) 1.25 [1.15, 1.37]
pre-op MRI and EEG
5 Good outcome by history of febrile 15 1368 Risk Ratio (M-H, Fixed, 95% CI) 1.09 [1.01, 1.17]
seizures (FS)
6 Good outcome by history of head in- 7 551 Risk Ratio (M-H, Fixed, 95% CI) 0.99 [0.86, 1.13]
jury (HI)

Informed decisions.


studies partici-
pants
7 Good outcome by presence of en- 4 274 Risk Ratio (M-H, Fixed, 95% CI) 0.67 [0.37, 1.21]
cephalomalacia
8 Good outcome by presence of focal 45 3529 Risk Ratio (M-H, Fixed, 95% CI) 0.90 [0.85, 0.95]
cortical dysplasia (FCD)/malformation
of cortical development (MCD)
9 Good outcome by presence of tu- 41 3357 Risk Ratio (M-H, Random, 95% CI) 1.23 [1.14, 1.32]
mour
9.1 > 1 year seizure-free 7 656 Risk Ratio (M-H, Random, 95% CI) 1.17 [0.97, 1.41]
9.2 Engel Class Scale 28 2199 Risk Ratio (M-H, Random, 95% CI) 1.23 [1.13, 1.33]
9.3 Other scale 6 502 Risk Ratio (M-H, Random, 95% CI) 1.35 [1.07, 1.70]
10 Good outcome by presence of vas- 19 1488 Risk Ratio (M-H, Fixed, 95% CI) 1.07 [0.94, 1.21]
cular malformation
11 Good outcome by unilateral or bi- 18 1414 Risk Ratio (M-H, Fixed, 95% CI) 1.14 [1.05, 1.24]
lateral interictal spikes
12 Good Outcome by Extent of Resec- 40 3013 Risk Ratio (M-H, Fixed, 95% CI) 1.41 [1.32, 1.50]
tion
12.1 > 1 year seizure free 9 640 Risk Ratio (M-H, Fixed, 95% CI) 2.00 [1.66, 2.41]

Informed decisions.


studies partici-
pants
13 Good outcome by extent of resec- 38 2887 Risk Ratio (M-H, Fixed, 95% CI) 1.48 [1.38, 1.58]
tion
13.1 Site of surgery: extratemporal on- 1 30 Risk Ratio (M-H, Fixed, 95% CI) 2.0 [0.76, 5.29]
ly
13.2 Site of surgery: temporal only 13 1266 Risk Ratio (M-H, Fixed, 95% CI) 1.11 [1.03, 1.20]
13.3 Site of surgery: temporal and ex- 24 1591 Risk Ratio (M-H, Fixed, 95% CI) 1.98 [1.77, 2.23]
tratemporal
14 Good outcome by side of surgical 37 2976 Risk Ratio (M-H, Random, 95% CI) 0.96 [0.91, 1.01]
resection
14.3 Other scale 5 279 Risk Ratio (M-H, Random, 95% CI) 0.94 [0.82, 1.07]
15 Good outcome by side of surgical 36 2933 Risk Ratio (M-H, Fixed, 95% CI) 0.94 [0.90, 0.98]
resection
15.1 Site of surgery: extratemporal on- 2 123 Risk Ratio (M-H, Fixed, 95% CI) 1.03 [0.76, 1.39]
ly
tratemporal
16 Good outcome by presence of post- 6 542 Risk Ratio (M-H, Random, 95% CI) 0.91 [0.68, 1.22]
operative discharges
17 Good outcome by the presence of 6 542 Risk Ratio (M-H, Fixed, 95% CI) 0.91 [0.79, 1.04]
postoperative discharges
tratemporal


Informed decisions.

Analysis 4.1. Comparison 4 Prognostic factors of a good outcome of surgery

(randomised and non-randomised evidence), Outcome 1 Good outcome by MRI results.
Study or subgroup Normal MRI Abnormal MRI Risk Ratio Weight Risk Ratio
Duchowny 1998 4/5 15/24 0.53% 1.28[0.75,2.19]
Erba 1992 6/10 31/36 1.38% 0.7[0.41,1.18]
Fujiwara 2012 6/15 16/26 1.2% 0.65[0.33,1.3]
Garcia 1994 10/20 26/31 2.08% 0.6[0.37,0.95]
Gilliam 1997a 11/19 42/59 2.09% 0.81[0.54,1.23]
Gilliam 1997b 2/3 20/30 0.37% 1[0.43,2.31]
Hallbook 2010 24/29 64/81 3.45% 1.05[0.86,1.28]
Jehi 2012 10/33 137/279 2.96% 0.62[0.36,1.05]
Kim 2009 49/96 45/70 5.31% 0.79[0.61,1.03]
Kim 2010b 7/24 7/16 0.86% 0.67[0.29,1.54]
Paolicchi 2000 18/35 26/40 2.48% 0.79[0.53,1.17]
Spencer 2005 22/40 157/225 4.84% 0.79[0.59,1.06]
Swartz 1992 5/5 22/29 0.77% 1.22[0.89,1.68]
Theodore 2012 8/15 18/26 1.34% 0.77[0.45,1.32]
Widdess-Walsh 2007 12/25 10/23 1.06% 1.1[0.59,2.05]
Yu 2012a 12/21 49/79 2.1% 0.92[0.61,1.39]
Yu 2012b 36/62 112/160 6.39% 0.83[0.66,1.05]
Subtotal (95% CI) 457 1234 39.2% 0.82[0.74,0.9]
Total events: 242 (Normal MRI), 797 (Abnormal MRI)
Heterogeneity: Tau2=0; Chi2=20.1, df=16(P=0.22); I2=20.39%
Test for overall effect: Z=4.02(P<0.0001)

Adam 1996 0/1 26/29 0.34% 0.28[0.03,3.13]
Alfstad 2011 5/8 20/40 0.68% 1.25[0.67,2.32]
Berkovic 1995 8/24 66/111 2.4% 0.56[0.31,1.01]
Chabardes 2005 7/8 34/40 1.16% 1.03[0.77,1.38]
Cossu 2005 16/42 77/123 4% 0.61[0.4,0.92]
Erickson 2005 8/16 49/63 2.03% 0.64[0.39,1.07]
Hartley 2002 2/4 18/31 0.42% 0.86[0.31,2.4]
Hartzfield 2008 5/15 19/41 1.04% 0.72[0.33,1.58]
Kim 2010a 34/111 41/66 5.25% 0.49[0.35,0.69]
Liang 2012 67/88 106/118 9.25% 0.85[0.74,0.97]
Liava 2012 5/10 34/42 1.34% 0.62[0.33,1.17]
Lorenzo 1995 7/26 10/22 1.11% 0.59[0.27,1.29]
Mihara 2004 150/195 71/87 10.03% 0.94[0.83,1.07]
O'Brien 2000 7/17 7/19 0.67% 1.12[0.49,2.53]
Perego 2009 2/6 27/31 0.89% 0.38[0.12,1.2]
Perry 2010 5/18 29/65 1.28% 0.62[0.28,1.38]
Radhakrishnan 1998 45/71 89/104 7.37% 0.74[0.61,0.9]
Sinclair 2003 5/12 52/65 1.65% 0.52[0.26,1.03]
Sindou 2006 8/13 77/87 2.04% 0.7[0.45,1.08]
Sola 2005 21/34 75/99 3.92% 0.82[0.61,1.09]
Wray 2012 2/6 34/46 0.8% 0.45[0.14,1.42]
Subtotal (95% CI) 725 1329 57.66% 0.74[0.69,0.8]
Favours abnormal MRI 0.05 0.2 1 5 20 Favours normal MRI

Informed decisions.

Study or subgroup Normal MRI Abnormal MRI Risk Ratio Weight Risk Ratio

4.1.3 Other scale
Boshuisen 2010 28/32 5/11 0.76% 1.93[0.99,3.73]
Goldstein 1996 5/14 10/19 0.87% 0.68[0.3,1.54]
Sakamoto 2009 0/4 19/32 0.52% 0.17[0.01,2.38]
Yang 2011 3/20 24/79 0.99% 0.49[0.17,1.48]
Subtotal (95% CI) 70 141 3.14% 0.84[0.53,1.32]

Total (95% CI) 1252 2704 100% 0.78[0.73,0.82]
Heterogeneity: Tau2=0; Chi2=68, df=41(P=0.01); I2=39.71%
Test for subgroup differences: Chi2=2.41, df=1 (P=0.3), I2=16.93%
Favours abnormal MRI 0.05 0.2 1 5 20 Favours normal MRI

Analysis 4.2. Comparison 4 Prognostic factors of a good outcome of surgery (randomised and
non-randomised evidence), Outcome 2 Good outcome by use of intracranial monitoring (IM).
Study or subgroup IM used IM not used Risk Ratio Weight Risk Ratio
Garcia 1994 8/14 28/37 3.7% 0.76[0.46,1.23]
Kim 2009 87/146 7/20 2.97% 1.7[0.92,3.14]
Kim 2010b 10/33 4/7 1.59% 0.53[0.23,1.21]
Spencer 2005 73/116 129/181 24.27% 0.88[0.75,1.04]
Tatum 2008 4/8 18/31 1.78% 0.86[0.4,1.83]
Theodore 2012 7/17 19/24 3.79% 0.52[0.28,0.95]
Subtotal (95% CI) 334 300 38.1% 0.88[0.76,1.02]
Total events: 189 (IM used), 205 (IM not used)

Adam 1996 5/8 21/22 2.7% 0.65[0.38,1.13]
Alfstad 2011 7/13 18/35 2.35% 1.05[0.58,1.9]
Chang 2009 8/13 24/31 3.42% 0.79[0.5,1.27]
Cossu 2008 62/96 15/17 6.14% 0.73[0.58,0.92]
Gelinas 2011 27/34 25/33 6.11% 1.05[0.81,1.36]
Grivas 2006 11/11 36/41 3.91% 1.1[0.93,1.3]
Lee 2008 34/65 4/6 1.76% 0.78[0.43,1.45]
Liava 2012 23/33 16/19 4.89% 0.83[0.61,1.11]
Morris 1998 14/21 12/15 3.37% 0.83[0.56,1.24]
O'Brien 2000 6/20 8/16 2.14% 0.6[0.26,1.38]
Perry 2010 16/34 18/49 3.55% 1.28[0.77,2.14]
Raabe 2012 15/25 42/51 6.65% 0.73[0.52,1.03]
Terra-Bustamante 2005a 11/15 16/20 3.3% 0.92[0.63,1.33]
Favours IM not used 0.01 0.1 1 10 100 Favours IM used

Informed decisions.

Study or subgroup IM used IM not used Risk Ratio Weight Risk Ratio
Wyllie 1998 15/30 66/90 7.95% 0.68[0.47,1]
Subtotal (95% CI) 418 445 58.24% 0.85[0.77,0.94]

4.2.3 Other Scale
Kan 2008 5/10 38/40 3.66% 0.53[0.28,0.98]
Subtotal (95% CI) 10 40 3.66% 0.53[0.28,0.98]

Total (95% CI) 762 785 100% 0.85[0.78,0.93]
Favours IM not used 0.01 0.1 1 10 100 Favours IM used

Analysis 4.3. Comparison 4 Prognostic factors of a good outcome of surgery (randomised and non-
randomised evidence), Outcome 3 Good outcome by presence of mesial temporal sclerosis (MTS).
Study or subgroup MTS present MTS absent Risk Ratio Weight Risk Ratio
Dagar 2011 23/26 66/86 3.03% 1.15[0.96,1.38]
Erba 1992 8/9 29/37 1.12% 1.13[0.85,1.51]
Gilliam 1997b 1/2 21/31 0.25% 0.74[0.18,3.01]
Hemb 2010 24/27 199/298 3.27% 1.33[1.14,1.56]
Jennum 1993 4/7 38/57 0.82% 0.86[0.44,1.67]
O'Brien 1996 27/31 9/15 1.2% 1.45[0.94,2.24]
Spencer 2005 128/171 47/86 6.19% 1.37[1.11,1.69]
Swartz 1992 12/15 15/19 1.31% 1.01[0.72,1.43]
Theodore 2012 16/23 10/18 1.11% 1.25[0.76,2.05]
Subtotal (95% CI) 311 647 18.32% 1.25[1.13,1.39]
Total events: 243 (MTS present), 434 (MTS absent)
Heterogeneity: Tau2=0; Chi2=6.24, df=8(P=0.62); I2=0%

Aaberg 2012 5/7 25/48 0.63% 1.37[0.8,2.36]
Alfstad 2011 12/18 13/30 0.97% 1.54[0.91,2.6]
Bell 2009 4/7 20/33 0.69% 0.94[0.47,1.89]
Berkovic 1995 49/85 25/50 3.12% 1.15[0.83,1.61]
Chabardes 2005 23/28 18/20 2.08% 0.91[0.73,1.14]
Cossu 2005 18/27 13/22 1.42% 1.13[0.73,1.75]
Cossu 2008 7/11 70/102 1.35% 0.93[0.58,1.48]
Delbeke 1996 16/27 7/11 0.98% 0.93[0.54,1.61]
Favours MTS absent 0.01 0.1 1 10 100 Favours MTS present

Informed decisions.

Study or subgroup MTS present MTS absent Risk Ratio Weight Risk Ratio
Donadio 2011 25/32 35/52 2.64% 1.16[0.89,1.51]
Erickson 2005 19/22 35/55 1.98% 1.36[1.05,1.76]
Georgakoulias 2008 15/21 21/23 1.98% 0.78[0.58,1.05]
Grivas 2006 22/31 15/21 1.77% 0.99[0.7,1.41]
Hartley 2002 8/13 12/22 0.88% 1.13[0.64,2]
Hartzfield 2008 4/10 20/46 0.71% 0.92[0.4,2.1]
Kanner 2009 60/63 26/37 3.24% 1.36[1.09,1.68]
Kilpatrick 1997 29/34 10/16 1.35% 1.36[0.91,2.04]
Li 1999 10/20 6/18 0.63% 1.5[0.68,3.29]
Mihara 2004 138/172 131/185 12.5% 1.13[1.01,1.28]
Miserocchi 2013 19/20 39/48 2.27% 1.17[0.99,1.38]
O'Brien 2000 0/1 14/35 0.15% 0.62[0.05,7.07]
Perego 2009 21/26 8/11 1.11% 1.11[0.74,1.67]
Perry 2010 3/7 31/76 0.52% 1.05[0.43,2.58]
Radhakrishnan 1998 63/74 71/101 5.94% 1.21[1.03,1.42]
Schramm 2011 133/186 19/21 3.38% 0.79[0.67,0.93]
Seymour 2012 85/158 63/143 6.55% 1.22[0.97,1.54]
Sinclair 2003 12/14 45/63 1.62% 1.2[0.92,1.56]
Sindou 2006 63/69 22/31 3.01% 1.29[1.02,1.63]
Wray 2012 5/9 31/43 1.06% 0.77[0.42,1.42]
Wyllie 1998 14/21 67/99 2.32% 0.99[0.71,1.37]
Zentner 1995 26/39 77/128 3.56% 1.11[0.85,1.44]
Subtotal (95% CI) 1269 1680 71.98% 1.13[1.07,1.2]

4.3.3 Other Scale
Goldstein 1996 1/5 14/28 0.42% 0.4[0.07,2.4]
Jeong 1999 74/85 4/8 0.72% 1.74[0.87,3.5]
Kan 2008 14/16 29/42 1.58% 1.27[0.96,1.67]
Mathern 1999 7/8 71/120 0.88% 1.48[1.09,2]
Sakamoto 2009 63/74 71/101 5.94% 1.21[1.03,1.42]
Yang 2011 18/28 1/8 0.15% 5.14[0.81,32.84]
Subtotal (95% CI) 216 307 9.7% 1.31[1.14,1.51]

Total (95% CI) 1796 2634 100% 1.17[1.12,1.23]
Favours MTS absent 0.01 0.1 1 10 100 Favours MTS present


Informed decisions.

non-randomised evidence), Outcome 4 Good outcome by Concordance of pre-op MRI and EEG.
Study or subgroup Concordant Discordant Risk Ratio Weight Risk Ratio
Garcia 1994 34/45 2/6 0.98% 2.27[0.72,7.11]
Gilliam 1997a 37/48 16/30 5.49% 1.45[1,2.09]
Gilliam 1997b 15/20 7/13 2.36% 1.39[0.79,2.45]
Kim 2009 68/127 26/39 11.09% 0.8[0.61,1.06]
Kim 2010b 12/29 2/11 0.81% 2.28[0.6,8.57]
Spencer 2005 111/153 90/143 25.93% 1.15[0.98,1.35]
Tatum 2008 20/27 2/12 0.77% 4.44[1.23,16.06]
Theodore 2012 16/24 10/17 3.26% 1.13[0.7,1.85]
Subtotal (95% CI) 473 271 50.7% 1.21[1.07,1.37]
Total events: 313 (Concordant), 155 (Discordant)

Brainer-Lima 1996 24/26 5/6 2.26% 1.11[0.76,1.61]
Chang 2009 4/5 28/39 1.77% 1.11[0.69,1.8]
Cukiert 2002 84/94 5/6 2.62% 1.07[0.74,1.54]
Dalmagro 2005 19/28 9/16 3.19% 1.21[0.73,1.99]
Janszky 2003a 47/64 13/20 5.52% 1.13[0.79,1.61]
Jaramillo-Betancur 2009 43/64 0/3 0.26% 5.35[0.4,71.97]
Kilpatrick 1997 37/48 2/2 1.31% 0.92[0.54,1.56]
Lei 2008 17/23 20/43 3.89% 1.59[1.06,2.38]
O'Brien 2000 7/18 7/18 1.95% 1[0.44,2.27]
Perego 2009 25/32 4/5 1.93% 0.98[0.61,1.57]
Sinclair 2003 45/58 12/19 5.04% 1.23[0.85,1.78]
Sola 2005 77/96 20/37 8.05% 1.48[1.08,2.03]
Subtotal (95% CI) 556 214 37.8% 1.27[1.11,1.46]

4.4.3 Other scale
Holmes 2000 33/76 21/50 7.06% 1.03[0.68,1.57]
Rossi 1994 28/38 9/21 3.23% 1.72[1.01,2.92]
Yang 2011 21/57 3/22 1.21% 2.7[0.89,8.16]
Subtotal (95% CI) 171 93 11.5% 1.4[1.02,1.93]
Heterogeneity: Tau2=0; Chi2=4, df=2(P=0.14); I2=49.95%

Total (95% CI) 1200 578 100% 1.25[1.15,1.37]
Test for subgroup differences: Chi2=0.86, df=1 (P=0.65), I2=0%
Favours Discordant 0.01 0.1 1 10 100 Favours Concordant


Informed decisions.

Analysis 4.5. Comparison 4 Prognostic factors of a good outcome of surgery (randomised

and non-randomised evidence), Outcome 5 Good outcome by history of febrile seizures (FS).
Study or subgroup History of FS No history of FS Risk Ratio Weight Risk Ratio
Holmes 1997 7/8 15/36 1.51% 2.1[1.32,3.35]
Kim 2009 17/30 77/136 7.73% 1[0.71,1.41]
Kim 2010b 6/11 8/29 1.22% 1.98[0.89,4.4]
Spencer 2005 67/86 134/210 21.62% 1.22[1.05,1.42]
Walz 2003 20/24 65/74 8.84% 0.95[0.78,1.16]
Subtotal (95% CI) 159 485 40.93% 1.18[1.05,1.32]
Total events: 117 (History of FS), 299 (No history of FS)

Adam 1996 14/16 12/14 3.55% 1.02[0.77,1.35]
Chabardes 2005 18/22 23/26 5.85% 0.92[0.73,1.18]
Grivas 2006 3/4 41/45 1.86% 0.82[0.46,1.46]
Hajek 2009 14/18 11/17 3.14% 1.2[0.78,1.85]
Perry 2010 3/5 31/78 1.04% 1.51[0.7,3.25]
Radhakrishnan 1998 45/57 89/118 16.1% 1.05[0.88,1.24]
Tezer 2008 60/73 30/36 11.16% 0.99[0.82,1.18]
Subtotal (95% CI) 220 411 49.96% 1.01[0.92,1.11]

4.5.3 Other scale
Jeong 1999 53/61 25/32 9.11% 1.11[0.9,1.37]
Subtotal (95% CI) 61 32 9.11% 1.11[0.9,1.37]
Test for overall effect: Z=1(P=0.32)

Total (95% CI) 440 928 100% 1.09[1.01,1.17]
Favours no history of FS 0.01 0.1 1 10 100 Favours history of FS

and non-randomised evidence), Outcome 6 Good outcome by history of head injury (HI).
Study or subgroup History of HI No history of HI Risk Ratio Weight Risk Ratio
Greiner 2011 0/1 42/53 2.33% 0.32[0.03,3.52]
Favours no history of HI 0.01 0.1 1 10 100 Favours history of HI

Informed decisions.

Study or subgroup History of HI No history of HI Risk Ratio Weight Risk Ratio

Walz 2003 31/38 54/60 32.08% 0.91[0.76,1.08]
Subtotal (95% CI) 39 113 34.41% 0.87[0.72,1.05]
Total events: 31 (History of HI), 96 (No history of HI)

Adam 1996 5/5 21/25 6.18% 1.11[0.82,1.49]
Terra-Bustamante 2005a 13/14 14/21 8.58% 1.39[1,1.95]
Tezer 2008 34/43 47/66 28.41% 1.11[0.89,1.38]
Subtotal (95% CI) 62 112 43.17% 1.17[0.99,1.37]

4.6.3 Other scale
Holmes 2000 6/24 48/102 14.01% 0.53[0.26,1.09]
Yang 2011 11/34 16/65 8.42% 1.31[0.69,2.51]
Subtotal (95% CI) 58 167 22.43% 0.83[0.51,1.33]

Total (95% CI) 159 392 100% 0.99[0.86,1.13]
Favours no history of HI 0.01 0.1 1 10 100 Favours history of HI

and non-randomised evidence), Outcome 7 Good outcome by presence of encephalomalacia.
Study or subgroup Encephaloma- Encephaloma- Risk Ratio Weight Risk Ratio
lacia present lacia absent
Hartzfield 2008 3/10 21/46 33.03% 0.66[0.24,1.78]
O'Brien 2000 3/7 11/29 18.84% 1.13[0.43,2.99]
Perry 2010 2/6 32/77 20.37% 0.8[0.25,2.56]
Subtotal (95% CI) 23 152 72.24% 0.82[0.45,1.5]
Total events: 8 (Encephalomalacia present), 64 (Encephalomalacia absent)

4.7.2 Other scale
Yang 2011 1/12 26/87 27.76% 0.28[0.04,1.87]
Subtotal (95% CI) 12 87 27.76% 0.28[0.04,1.87]
Favours encephalomalacia absent 0.01 0.1 1 10 100 Favours encephalomalacia present

Informed decisions.

Study or subgroup Encephaloma- Encephaloma- Risk Ratio Weight Risk Ratio

lacia present lacia absent

Total (95% CI) 35 239 100% 0.67[0.37,1.21]
Favours encephalomalacia absent 0.01 0.1 1 10 100 Favours encephalomalacia present

Analysis 4.8. Comparison 4 Prognostic factors of a good outcome of surgery
(randomised and non-randomised evidence), Outcome 8 Good outcome by presence
of focal cortical dysplasia (FCD)/malformation of cortical development (MCD).
Study or subgroup FCD/MCD FCD/MCD Risk Ratio Weight Risk Ratio
present absent
Dagar 2011 13/18 76/94 2.89% 0.89[0.66,1.21]
Duchowny 1998 9/21 7/10 1.12% 0.61[0.32,1.16]
Fujiwara 2012 16/34 7/10 1.28% 0.67[0.39,1.15]
Gilliam 1997b 5/12 17/21 1.46% 0.51[0.26,1.04]
Greiner 2011 17/21 25/33 2.3% 1.07[0.8,1.42]
Hallbook 2010 31/43 57/67 5.27% 0.85[0.69,1.05]
Hemb 2010 86/120 137/205 11.96% 1.07[0.92,1.24]
Paolicchi 2000 22/42 22/33 2.91% 0.79[0.54,1.14]
Subtotal (95% CI) 311 473 29.19% 0.92[0.84,1.02]
Total events: 199 (FCD/MCD present), 348 (FCD/MCD absent)

Aaberg 2012 10/21 20/33 1.84% 0.79[0.46,1.33]
Battaglia 2006 7/16 25/29 2.1% 0.51[0.29,0.9]
Blount 2004 2/6 10/24 0.47% 0.8[0.23,2.73]
Cossu 2005 41/72 52/93 5.37% 1.02[0.78,1.33]
Cossu 2008 54/88 23/25 4.24% 0.67[0.54,0.82]
Cukiert 2002 5/6 84/94 1.19% 0.93[0.65,1.34]
Dalmagro 2005 15/22 13/21 1.57% 1.1[0.71,1.71]
Donadio 2011 11/15 49/69 2.07% 1.03[0.73,1.45]
Elsharkawy 2008a 28/68 93/150 6.86% 0.66[0.49,0.91]
Erickson 2005 3/5 51/72 0.78% 0.85[0.41,1.76]
Georgakoulias 2008 3/3 33/41 0.69% 1.1[0.73,1.64]
Hajek 2009 9/14 16/21 1.51% 0.84[0.53,1.33]
Hamiwka 2005 12/31 5/7 0.96% 0.54[0.28,1.03]
Kilpatrick 1997 1/2 38/48 0.36% 0.63[0.16,2.54]
Kim 2010a 43/103 32/74 4.4% 0.97[0.68,1.37]
Lee 2011 8/9 29/31 1.54% 0.95[0.74,1.22]
Li 1997 5/7 31/44 1.01% 1.01[0.61,1.68]
Favours FCD/MCD absent 0.01 0.1 1 10 100 Favours FCD/MCD present

Informed decisions.

Study or subgroup FCD/MCD FCD/MCD Risk Ratio Weight Risk Ratio

present absent
Li 1999 4/9 12/29 0.67% 1.07[0.46,2.51]
Liava 2012 22/28 17/24 2.16% 1.11[0.8,1.53]
Miserocchi 2013 30/32 28/36 3.12% 1.21[0.99,1.47]
Morris 1998 9/13 17/23 1.45% 0.94[0.61,1.45]
O'Brien 2000 2/3 12/33 0.24% 1.83[0.73,4.59]
Perego 2009 2/3 27/34 0.52% 0.84[0.37,1.9]
Perry 2010 16/40 18/43 2.05% 0.96[0.57,1.6]
Raabe 2012 2/3 55/73 0.51% 0.88[0.39,1.99]
Sinclair 2003 7/12 50/65 1.84% 0.76[0.46,1.25]
Sindou 2006 3/5 82/95 0.97% 0.7[0.34,1.43]
Tripathi 2008 16/26 13/29 1.45% 1.37[0.83,2.28]
Urbach 2007 11/16 14/26 1.26% 1.28[0.79,2.07]
Wray 2012 2/5 34/47 0.77% 0.55[0.19,1.64]
Wyllie 1998 16/31 65/89 3.97% 0.71[0.49,1.02]
Subtotal (95% CI) 741 1602 60.83% 0.88[0.82,0.96]

4.8.3 Other scale
Boesebeck 2007 10/26 23/55 1.75% 0.92[0.52,1.64]
Kan 2008 8/13 35/45 1.86% 0.79[0.5,1.25]
Mathern 1999 35/54 43/74 4.29% 1.12[0.85,1.47]
Sakamoto 2009 1/1 18/35 0.23% 1.46[0.62,3.45]
Yang 2011 6/37 21/62 1.86% 0.48[0.21,1.08]
Subtotal (95% CI) 131 271 9.98% 0.91[0.73,1.13]

Total (95% CI) 1183 2346 100% 0.9[0.85,0.95]
Favours FCD/MCD absent 0.01 0.1 1 10 100 Favours FCD/MCD present

and non-randomised evidence), Outcome 9 Good outcome by presence of tumour.
Study or subgroup Tumour Tumour absent Risk Ratio Weight Risk Ratio
present
n/N n/N M-H, Random, 95% CI M-H, Random, 95% CI
Dagar 2011 14/18 75/94 3.51% 0.97[0.75,1.27]
Duchowny 1998 7/7 9/24 1.44% 2.47[1.45,4.2]
Erba 1992 20/26 17/20 3.35% 0.9[0.68,1.2]
Gilliam 1997b 10/13 12/20 1.76% 1.28[0.8,2.04]
Favours tumour absent 0.01 0.1 1 10 100 Favours tumour present

Informed decisions.

present
Hemb 2010 18/22 205/303 4.29% 1.21[0.98,1.49]
Paolicchi 2000 17/25 27/50 2.42% 1.26[0.87,1.83]
Swartz 1992 5/6 22/28 2.14% 1.06[0.71,1.59]
Subtotal (95% CI) 117 539 18.91% 1.17[0.97,1.41]
Total events: 91 (Tumour present), 367 (Tumour absent)
Heterogeneity: Tau2=0.03; Chi2=13.11, df=6(P=0.04); I2=54.24%

Aaberg 2012 1/5 29/49 0.16% 0.34[0.06,1.98]
Battaglia 2006 3/12 29/33 0.49% 0.28[0.11,0.76]
Cossu 2005 17/30 76/135 2.64% 1.01[0.71,1.42]
Cossu 2008 31/35 46/78 4.16% 1.5[1.21,1.87]
Dalmagro 2005 5/6 23/37 1.93% 1.34[0.87,2.08]
Elsharkawy 2008a 50/79 71/139 3.96% 1.24[0.98,1.57]
Erickson 2005 10/16 44/61 2.11% 0.87[0.57,1.31]
Grivas 2006 6/7 31/45 2.51% 1.24[0.87,1.78]
Hamiwka 2005 5/7 12/31 1.05% 1.85[0.97,3.52]
Kilpatrick 1997 6/10 33/40 1.46% 0.73[0.43,1.23]
Kim 2010a 6/10 69/167 1.41% 1.45[0.85,2.49]
Lee 2011 28/30 9/10 4.05% 1.04[0.83,1.3]
Li 1997 24/32 12/19 2.21% 1.19[0.8,1.77]
Li 1999 5/10 11/18 0.86% 0.82[0.4,1.68]
Liava 2012 10/12 29/40 2.93% 1.15[0.84,1.58]
Miserocchi 2013 32/35 26/33 4.41% 1.16[0.95,1.42]
O'Brien 2000 3/8 11/28 0.47% 0.95[0.35,2.61]
Perego 2009 6/7 23/30 2.5% 1.12[0.78,1.6]
Perry 2010 12/18 22/65 1.73% 1.97[1.23,3.16]
Sinclair 2003 18/21 39/56 3.8% 1.23[0.96,1.57]
Sindou 2006 9/10 76/90 4.09% 1.07[0.85,1.33]
Sperling 1992 15/16 26/35 3.98% 1.26[1,1.59]
Urbach 2007 8/13 17/29 1.46% 1.05[0.62,1.78]
Wray 2012 20/22 16/30 2.51% 1.7[1.19,2.44]
Wyllie 1998 36/44 45/76 3.98% 1.38[1.09,1.74]
Zentner 1995 50/73 53/94 3.93% 1.21[0.96,1.54]
Zentner 1996 12/15 18/41 1.99% 1.82[1.19,2.8]
Subtotal (95% CI) 601 1598 70.08% 1.23[1.13,1.33]

4.9.3 Other scale
Boesebeck 2007 11/18 22/63 1.58% 1.75[1.06,2.88]
Goldstein 1996 8/11 7/22 0.89% 2.29[1.12,4.65]
Kan 2008 13/16 30/42 3.08% 1.14[0.84,1.54]
Mathern 1999 6/12 72/116 1.24% 0.81[0.45,1.44]
Rossi 1994 20/24 47/79 3.67% 1.4[1.09,1.81]
Yang 2011 3/7 24/92 0.56% 1.64[0.65,4.13]
Subtotal (95% CI) 88 414 11.01% 1.35[1.07,1.7]

Informed decisions.

present

Total (95% CI) 806 2551 100% 1.23[1.14,1.32]
Heterogeneity: Tau2=0.02; Chi2=67.33, df=40(P=0); I2=40.59%

non-randomised evidence), Outcome 10 Good outcome by presence of vascular malformation.
Study or subgroup Malforma- Malforma- Risk Ratio Weight Risk Ratio
tion present tion absent
Erba 1992 2/2 35/44 2.95% 1.06[0.62,1.79]
Subtotal (95% CI) 2 44 2.95% 1.06[0.62,1.79]
Total events: 2 (Malformation present), 35 (Malformation absent)

Aaberg 2012 8/9 22/45 4.87% 1.82[1.25,2.65]
Battaglia 2006 3/4 29/41 3.42% 1.06[0.58,1.93]
Cossu 2008 3/3 74/110 3.44% 1.3[0.88,1.93]
Dalmagro 2005 2/2 26/41 2.35% 1.32[0.76,2.3]
Elsharkawy 2008a 15/24 106/194 15.49% 1.14[0.82,1.6]
Erickson 2005 6/8 47/69 6.48% 1.1[0.72,1.7]
Kilpatrick 1997 3/4 36/46 3.82% 0.96[0.53,1.72]
Li 1997 5/8 31/43 6.46% 0.87[0.49,1.53]
Li 1999 2/5 14/33 2.45% 0.94[0.3,2.96]
Perry 2010 0/3 34/80 2.16% 0.29[0.02,3.97]
Urbach 2007 3/4 22/38 2.78% 1.3[0.69,2.43]
Wray 2012 2/2 34/50 2.54% 1.23[0.72,2.11]
Subtotal (95% CI) 80 893 59.18% 1.14[0.98,1.34]

4.10.3 Other Scale
Boesebeck 2007 2/8 31/73 4.06% 0.59[0.17,2.01]
Kan 2008 4/5 39/53 4.46% 1.09[0.68,1.73]
Mathern 1999 12/21 66/107 14.38% 0.93[0.62,1.38]
Rossi 1994 13/20 54/83 13.92% 1[0.7,1.43]
Yang 2011 1/3 26/96 1.05% 1.23[0.24,6.3]
Subtotal (95% CI) 57 412 37.87% 0.94[0.74,1.2]
Favours malformation absent 0.01 0.1 1 10 100 Favours malformation present

Informed decisions.

Study or subgroup Malforma- Malforma- Risk Ratio Weight Risk Ratio

tion present tion absent

Total (95% CI) 139 1349 100% 1.07[0.94,1.21]
Favours malformation absent 0.01 0.1 1 10 100 Favours malformation present

non-randomised evidence), Outcome 11 Good outcome by unilateral or bilateral interictal spikes.
Study or subgroup Unilater- Bilateral Spikes Risk Ratio Weight Risk Ratio
al spikes
Chee 1993 30/33 2/5 1.01% 2.27[0.77,6.69]
Greiner 2011 29/38 13/16 5.32% 0.94[0.7,1.26]
Kim 2009 41/71 53/95 13.19% 1.04[0.79,1.35]
Kim 2010b 10/28 4/12 1.63% 1.07[0.42,2.75]
Tatum 2008 16/27 6/12 2.42% 1.19[0.62,2.26]
Walz 2003 39/42 44/53 11.32% 1.12[0.96,1.3]
Weinand 1992 52/82 5/7 2.68% 0.89[0.54,1.46]
Subtotal (95% CI) 321 200 37.58% 1.08[0.94,1.24]
Total events: 217 (Unilateral spikes), 127 (Bilateral Spikes)

Dalmagro 2005 15/23 13/19 4.14% 0.95[0.62,1.46]
Erickson 2005 45/60 6/13 2.87% 1.63[0.89,2.98]
Jayakar 2008 45/75 7/26 3.03% 2.23[1.15,4.31]
Lee 2008 10/16 27/50 3.81% 1.16[0.73,1.83]
Madhavan 2007 24/35 11/31 3.4% 1.93[1.14,3.27]
Remi 2011 67/78 71/76 20.93% 0.92[0.83,1.02]
Subtotal (95% CI) 287 215 38.18% 1.19[1.04,1.36]
Heterogeneity: Tau2=0; Chi2=31.24, df=5(P<0.0001); I2=83.99%

4.11.3 Other scale
Boshuisen 2010 16/19 17/24 4.37% 1.19[0.86,1.64]
Goldstein 1996 9/21 6/12 2.22% 0.86[0.4,1.82]
Holmes 2000 20/26 34/100 4.08% 2.26[1.6,3.19]
Lee 2006 23/29 14/22 4.63% 1.25[0.86,1.8]
Rossi 1994 13/29 73/109 8.93% 0.67[0.44,1.02]
Subtotal (95% CI) 124 267 24.24% 1.16[0.97,1.39]
Favours bilateral spikes 0.1 0.2 0.5 1 2 5 10 Favours unilateral spikes

Informed decisions.

Study or subgroup Unilater- Bilateral Spikes Risk Ratio Weight Risk Ratio
al spikes
Heterogeneity: Tau2=0; Chi2=21.67, df=4(P=0); I2=81.54%

Total (95% CI) 732 682 100% 1.14[1.05,1.24]
Favours bilateral spikes 0.1 0.2 0.5 1 2 5 10 Favours unilateral spikes

and non-randomised evidence), Outcome 12 Good Outcome by Extent of Resection.
Study or subgroup Complete Less Complete Risk Ratio Weight Risk Ratio
4.12.1 > 1 year seizure free
Awad 1991 17/18 10/29 1.14% 2.74[1.64,4.58]
Fujiwara 2012 15/26 8/18 1.41% 1.3[0.7,2.39]
Jennum 1993 36/50 6/14 1.4% 1.68[0.9,3.15]
Kim 2009 77/111 17/55 3.38% 2.24[1.48,3.4]
Kim 2010b 12/24 2/16 0.36% 4[1.03,15.53]
Paolicchi 2000 37/49 7/26 1.36% 2.8[1.46,5.39]
Walz 2003 72/82 9/11 2.36% 1.07[0.8,1.43]
Widdess-Walsh 2007 15/22 5/19 0.8% 2.59[1.16,5.79]
Wyler 1995 25/36 13/34 1.99% 1.82[1.12,2.93]
Subtotal (95% CI) 418 222 14.19% 2[1.66,2.41]
Total events: 306 (Complete), 77 (Less Complete)

Arruda 1996 25/37 28/37 4.17% 0.89[0.67,1.19]
Babini 2013 10/10 16/30 1.29% 1.79[1.26,2.56]
Blount 2004 12/26 0/4 0.13% 4.63[0.32,66.24]
Brainer-Lima 1996 25/25 4/7 1.02% 1.74[0.94,3.22]
Chang 2009 32/42 0/2 0.14% 4.53[0.36,57.27]
Choi 2004a 20/22 7/13 1.31% 1.69[1,2.84]
Cossu 2005 70/115 23/50 4.77% 1.32[0.95,1.85]
Cossu 2008 57/72 20/41 3.79% 1.62[1.16,2.27]
Dalmagro 2005 11/18 17/25 2.12% 0.9[0.57,1.42]
Hamiwka 2005 24/29 1/9 0.23% 7.45[1.16,47.62]
Jayakar 2008 35/63 9/38 1.67% 2.35[1.27,4.33]
Kanner 2009 23/25 63/75 4.69% 1.1[0.94,1.28]
Kloss 2002 21/26 5/30 0.69% 4.85[2.13,11.02]
Lee 2011 32/32 5/8 1.29% 1.61[0.96,2.72]
Li 1997 35/41 4/10 0.96% 2.13[0.99,4.61]
Madhavan 2007 26/39 11/31 1.82% 1.88[1.11,3.17]
Favours Less Complete 0.01 0.1 1 10 100 Favours Complete

Informed decisions.

Study or subgroup Complete Less Complete Risk Ratio Weight Risk Ratio
Miserocchi 2013 54/63 4/5 1.1% 1.07[0.68,1.68]
Morris 1998 20/28 6/8 1.39% 0.95[0.6,1.51]
O'Brien 2000 4/4 6/11 0.57% 1.66[0.91,3.02]
Paglioli 2006 58/80 58/81 8.58% 1.01[0.84,1.23]
Raabe 2012 10/26 47/50 4.79% 0.41[0.25,0.67]
Sagher 2012 43/51 38/45 6.01% 1[0.84,1.19]
Schramm 2011 62/74 84/125 9.3% 1.25[1.06,1.46]
Stavrou 2008 14/18 23/35 2.32% 1.18[0.84,1.67]
Tanriverdi 2010 81/123 79/133 11.3% 1.11[0.92,1.34]
Yeon 2009 21/27 29/33 3.88% 0.89[0.7,1.12]
Zentner 1996 14/21 3/9 0.62% 2[0.76,5.29]
Subtotal (95% CI) 1216 973 80.59% 1.29[1.21,1.39]

4.12.3 Other scale
Lackmayer 2013 5/9 27/36 1.61% 0.74[0.4,1.37]
Rossi 1994 54/63 13/40 2.37% 2.64[1.67,4.17]
Sakamoto 2009 4/10 15/26 1.24% 0.69[0.3,1.59]
Subtotal (95% CI) 82 102 5.21% 1.59[1.15,2.2]

Total (95% CI) 1716 1297 100% 1.41[1.32,1.5]
Test for subgroup differences: Chi2=19.07, df=1 (P<0.0001), I2=89.51%
Favours Less Complete 0.01 0.1 1 10 100 Favours Complete

and non-randomised evidence), Outcome 13 Good outcome by extent of resection.
Study or subgroup Complete Less com- Risk Ratio Weight Risk Ratio
resection plete resection
4.13.1 Site of surgery: extratemporal only
Zentner 1996 14/21 3/9 0.68% 2[0.76,5.29]
Subtotal (95% CI) 21 9 0.68% 2[0.76,5.29]
Total events: 14 (Complete resection), 3 (Less complete resection)

4.13.2 Site of surgery: temporal only
Arruda 1996 25/37 28/37 4.56% 0.89[0.67,1.19]
Choi 2004a 20/22 7/13 1.43% 1.69[1,2.84]
Favours less complete resection 0.01 0.1 1 10 100 Favours complete resection

Informed decisions.

Study or subgroup Complete Less com- Risk Ratio Weight Risk Ratio
resection plete resection
Kanner 2009 23/25 63/75 5.13% 1.1[0.94,1.28]
Lackmayer 2013 5/9 27/36 1.76% 0.74[0.4,1.37]
Miserocchi 2013 32/32 5/8 1.41% 1.61[0.96,2.72]
Paglioli 2006 54/63 4/5 1.21% 1.07[0.68,1.68]
Sagher 2012 58/80 58/81 9.39% 1.01[0.84,1.23]
Sakamoto 2009 43/51 38/45 6.58% 1[0.84,1.19]
Sarkis 2012 4/10 15/26 1.36% 0.69[0.3,1.59]
Schramm 2011 32/53 1/10 0.27% 6.04[0.93,39.26]
Tanriverdi 2010 62/74 84/125 10.18% 1.25[1.06,1.46]
Walz 2003 81/123 79/133 12.37% 1.11[0.92,1.34]
Wyler 1995 72/82 9/11 2.59% 1.07[0.8,1.43]
Subtotal (95% CI) 661 605 58.22% 1.11[1.03,1.2]

4.13.3 Site of surgery: temporal and extratemporal
Awad 1991 17/18 10/29 1.25% 2.74[1.64,4.58]
Babini 2013 10/10 16/30 1.41% 1.79[1.26,2.56]
Blount 2004 12/26 0/4 0.14% 4.63[0.32,66.24]
Brainer-Lima 1996 25/25 4/7 1.12% 1.74[0.94,3.22]
Chang 2009 32/42 0/2 0.15% 4.53[0.36,57.27]
Cossu 2005 70/115 23/50 5.22% 1.32[0.95,1.85]
Cossu 2008 57/72 20/41 4.15% 1.62[1.16,2.27]
Fujiwara 2012 15/26 8/18 1.54% 1.3[0.7,2.39]
Hamiwka 2005 24/29 1/9 0.25% 7.45[1.16,47.62]
Jayakar 2008 35/63 9/38 1.83% 2.35[1.27,4.33]
Jennum 1993 36/50 6/14 1.53% 1.68[0.9,3.15]
Kim 2009 77/111 17/55 3.7% 2.24[1.48,3.4]
Kim 2010b 12/24 2/16 0.39% 4[1.03,15.53]
Kloss 2002 21/26 5/30 0.76% 4.85[2.13,11.02]
Li 1997 35/41 4/10 1.05% 2.13[0.99,4.61]
Madhavan 2007 26/39 11/31 2% 1.88[1.11,3.17]
Morris 1998 20/28 6/8 1.52% 0.95[0.6,1.51]
O'Brien 2000 4/4 6/11 0.62% 1.66[0.91,3.02]
Paolicchi 2000 37/49 7/26 1.49% 2.8[1.46,5.39]
Rossi 1994 54/63 13/40 2.59% 2.64[1.67,4.17]
Stavrou 2008 14/18 23/35 2.54% 1.18[0.84,1.67]
Widdess-Walsh 2007 15/22 5/19 0.87% 2.59[1.16,5.79]
Yeon 2009 21/27 29/33 4.25% 0.89[0.7,1.12]
Subtotal (95% CI) 1007 584 41.09% 1.98[1.77,2.23]

Total (95% CI) 1689 1198 100% 1.48[1.38,1.58]
Test for subgroup differences: Chi2=67.21, df=1 (P<0.0001), I2=97.02%
Favours less complete resection 0.01 0.1 1 10 100 Favours complete resection

Informed decisions.


and non-randomised evidence), Outcome 14 Good outcome by side of surgical resection.
Study or subgroup Left side Right side Risk Ratio Weight Risk Ratio
resection resection
Jack 1992 22/27 18/23 2.64% 1.04[0.79,1.38]
Morino 2009 27/31 28/31 4.76% 0.96[0.81,1.15]
Tatum 2008 12/23 10/16 0.86% 0.83[0.48,1.44]
Theodore 2012 15/23 11/18 1.11% 1.07[0.66,1.71]
Walz 2003 54/61 31/37 5.06% 1.06[0.89,1.25]
Subtotal (95% CI) 165 125 14.42% 1.01[0.91,1.13]
Total events: 130 (Left side resection), 98 (Right side resection)

Adam 1996 13/15 13/15 2.64% 1[0.76,1.32]
Babini 2013 12/13 14/17 2.79% 1.12[0.86,1.47]
Battaglia 2006 17/26 15/19 1.75% 0.83[0.58,1.19]
Brainer-Lima 1996 14/15 15/17 3.7% 1.06[0.85,1.32]
Cascino 1995 63/92 50/73 3.96% 1[0.81,1.23]
Chabardes 2005 14/20 27/28 2.44% 0.73[0.54,0.98]
Chang 2009 14/20 18/23 1.79% 0.89[0.62,1.28]
Choi 2004a 17/22 10/13 1.67% 1[0.69,1.46]
Cossu 2008 37/56 40/57 3.07% 0.94[0.73,1.21]
Dalmagro 2005 13/21 15/22 1.26% 0.91[0.58,1.41]
Georgakoulias 2008 20/23 21/26 3.2% 1.08[0.84,1.38]
Grivas 2006 22/23 25/29 5.01% 1.11[0.94,1.31]
Hartzfield 2008 12/23 12/27 0.78% 1.17[0.66,2.09]
Janszky 2003a 23/33 20/34 1.78% 1.18[0.83,1.7]
Kilpatrick 1997 19/27 20/23 2.49% 0.81[0.6,1.08]
Lee 2008 22/34 16/37 1.24% 1.5[0.96,2.33]
Liava 2012 11/19 28/33 1.43% 0.68[0.45,1.03]
Miserocchi 2013 31/34 27/34 4.15% 1.15[0.94,1.4]
Perego 2009 11/13 18/24 2.08% 1.13[0.81,1.56]
Prevedello 2000 24/41 29/43 2.04% 0.87[0.62,1.21]
Schramm 2011 84/110 68/97 5.11% 1.09[0.92,1.29]
Seymour 2012 68/155 80/128 3.64% 0.7[0.56,0.88]
Sola 2005 35/74 39/63 2.28% 0.76[0.56,1.04]
Tanriverdi 2010 75/132 84/124 4.37% 0.84[0.69,1.02]
Tezer 2008 40/51 50/58 4.79% 0.91[0.76,1.09]
Zentner 1995 44/68 59/99 3.32% 1.09[0.85,1.38]
Subtotal (95% CI) 1206 1201 75.57% 0.96[0.89,1.02]

4.14.3 Other scale
Engman 2004 16/25 19/29 1.52% 0.98[0.66,1.45]
Favours right resection 0.01 0.1 1 10 100 Favours left resection

Informed decisions.

resection resection
Jeong 1999 35/42 43/51 4.71% 0.99[0.83,1.18]
Lackmayer 2013 14/22 18/23 1.61% 0.81[0.55,1.19]
Lee 2006 12/20 25/31 1.51% 0.74[0.5,1.11]
Sakamoto 2009 11/20 8/16 0.66% 1.1[0.59,2.07]
Subtotal (95% CI) 129 150 10% 0.94[0.82,1.07]

Total (95% CI) 1500 1476 100% 0.96[0.91,1.01]

and non-randomised evidence), Outcome 15 Good outcome by side of surgical resection.
4.15.1 Site of surgery: extratemporal only
Lee 2008 22/34 16/37 1.48% 1.5[0.96,2.33]
Liava 2012 11/19 28/33 1.98% 0.68[0.45,1.03]
Subtotal (95% CI) 53 70 3.46% 1.03[0.76,1.39]
Total events: 33 (Left side), 44 (Right side)

Adam 1996 13/15 13/15 1.26% 1[0.76,1.32]
Battaglia 2006 17/26 15/19 1.68% 0.83[0.58,1.19]
Cascino 1995 63/92 50/73 5.4% 1[0.81,1.23]
Chabardes 2005 14/20 27/28 2.18% 0.73[0.54,0.98]
Choi 2004a 17/22 10/13 1.22% 1[0.69,1.46]
Engman 2004 16/25 19/29 1.7% 0.98[0.66,1.45]
Georgakoulias 2008 20/23 21/26 1.91% 1.08[0.84,1.38]
Grivas 2006 22/23 25/29 2.14% 1.11[0.94,1.31]
Hartzfield 2008 12/23 12/27 1.07% 1.17[0.66,2.09]
Jack 1992 22/27 18/23 1.88% 1.04[0.79,1.38]
Janszky 2003a 28/46 32/38 3.39% 0.72[0.55,0.95]
Jeong 1999 35/42 43/51 3.76% 0.99[0.83,1.18]
Kilpatrick 1997 19/27 20/23 2.09% 0.81[0.6,1.08]
Lackmayer 2013 14/22 18/23 1.7% 0.81[0.55,1.19]
Lee 2006 12/20 25/31 1.9% 0.74[0.5,1.11]
Miserocchi 2013 31/34 27/34 2.61% 1.15[0.94,1.4]
Morino 2009 27/31 28/31 2.71% 0.96[0.81,1.15]

Informed decisions.

Perego 2009 11/13 18/24 1.22% 1.13[0.81,1.56]
Prevedello 2000 24/41 29/43 2.74% 0.87[0.62,1.21]
Sakamoto 2009 11/20 8/16 0.86% 1.1[0.59,2.07]
Schramm 2011 84/110 68/97 7% 1.09[0.92,1.29]
Seymour 2012 68/155 80/128 8.48% 0.7[0.56,0.88]
Sola 2005 35/74 39/63 4.08% 0.76[0.56,1.04]
Tanriverdi 2010 75/132 84/124 8.39% 0.84[0.69,1.02]
Tatum 2008 12/23 10/16 1.14% 0.83[0.48,1.44]
Tezer 2008 40/51 50/58 4.53% 0.91[0.76,1.09]
Theodore 2012 15/23 11/18 1.19% 1.07[0.66,1.71]
Walz 2003 54/61 31/37 3.74% 1.06[0.89,1.25]
Zentner 1995 44/68 59/99 4.65% 1.09[0.85,1.38]
Subtotal (95% CI) 1322 1270 88.54% 0.93[0.89,0.98]

Babini 2013 12/13 14/17 1.17% 1.12[0.86,1.47]
Brainer-Lima 1996 14/15 15/17 1.36% 1.06[0.85,1.32]
Chang 2009 14/20 18/23 1.62% 0.89[0.62,1.28]
Cossu 2008 37/56 40/57 3.84% 0.94[0.73,1.21]
Subtotal (95% CI) 104 114 8% 0.98[0.84,1.14]

Total (95% CI) 1479 1454 100% 0.94[0.9,0.98]

non-randomised evidence), Outcome 16 Good outcome by presence of postoperative discharges.
Study or subgroup Discharges Discharges Risk Ratio Weight Risk Ratio
present absent
Jennum 1993 31/38 11/26 15.08% 1.93[1.2,3.1]
Widdess-Walsh 2007 7/25 12/20 9.88% 0.47[0.23,0.96]
Subtotal (95% CI) 63 46 24.96% 0.97[0.24,3.93]
Total events: 38 (Discharges present), 23 (Discharges absent)

Favours discharges absent 0.01 0.1 1 10 100 Favours discharges present

Informed decisions.


present absent
Janszky 2003b 13/22 92/116 18.09% 0.75[0.52,1.07]
Miserocchi 2013 13/20 45/48 18.92% 0.69[0.5,0.96]
Radhakrishnan 1998 41/58 93/117 22.54% 0.89[0.74,1.07]
Wray 2012 27/37 9/15 15.49% 1.22[0.77,1.92]
Subtotal (95% CI) 137 296 75.04% 0.85[0.7,1.03]

Total (95% CI) 200 342 100% 0.91[0.68,1.22]

Analysis 4.17. Comparison 4 Prognostic factors of a good outcome of surgery (randomised and non-
randomised evidence), Outcome 17 Good outcome by the presence of postoperative discharges.
present absent
Janszky 2003b 13/22 92/116 18.73% 0.75[0.52,1.07]
Miserocchi 2013 41/58 93/117 39.35% 0.89[0.74,1.07]
Radhakrishnan 1998 13/20 45/48 16.9% 0.69[0.5,0.96]
Subtotal (95% CI) 100 281 74.97% 0.81[0.7,0.94]
Heterogeneity: Tau2=0; Chi2=2, df=2(P=0.37); I2=0%

Jennum 1993 27/37 9/15 8.18% 1.22[0.77,1.92]
Widdess-Walsh 2007 7/25 12/20 8.51% 0.47[0.23,0.96]
Wray 2012 31/38 11/26 8.34% 1.93[1.2,3.1]
Subtotal (95% CI) 100 61 25.03% 1.2[0.89,1.61]

Total (95% CI) 200 342 100% 0.91[0.79,1.04]

Informed decisions.

ADDITIONAL TABLES

Table 1. Study characteristics and participant demographics for 178 included studies
Author Outcome Scale used Study design Partici- Good out- Males Site of Surgery Age at Dura- Follow-up
(year) pants come surgery tion of (min max)
Library
Cochrane
(min max) epilepsy months1
years1 (min max)
years1
Aaberg 2012 Engel Class Scale Retrospective 54 30 32 Temporal and extratem- 0.5 to 16 NA > 24
poral
Better health.
Informed decisions.
Trusted evidence.
Adam 1996 Engel Class Scale Prospective 30 26 11 Temporal only 18 to 44 NA 12 to 44
Adelson > 1 year seizure-free Retrospective 33 23 19 Temporal and extratem- 0 to 17 NA 18 to 72

1992 poral
Alfstad 2011 Engel Class Scale Retrospective 48 25 26 Temporal and extratem- 24.6 (NA) 14.5 (NA) > 24
poral
Althausen > 1 year seizure-free Retrospective 61 45 29 Extratemporal only 14.5 (12.0) 11.9 (10.8) 13 to 233
2013
Arruda 1996 Engel Class Scale Retrospective 74 53 34 Temporal only 32.1 (10.5) NA 33.4 (13.1)
Awad 1991 > 1 year seizure-free Not stated 47 27 38 Temporal and extratem- 2 to 45 0 to 29 20 to 114
poral
Babini 2013 Engel Class Scale Retrospective 30 26 20 Temporal and extratem- 3 to 18 0 to 16 12 to 204
poral
Battaglia Engel Class Scale Not stated 45 32 29 Temporal only 0 to 7 0 to 5 24 to 179

2006
Baumann Engel Class Scale Retrospective 168 118 90 Temporal and extratem- 1 to 71 0 to 41 12 to 39

2007 poral
Bautista > 1 year seizure-free Retrospective 43 28 29 Temporal and extratem- 29.6 (10.9) 2 to 48 > 12
2003 poral
Bell 2009 Engel Class Scale Retrospective 40 24 17 Temporal only 13 to 62 1 to 36 18 to 126
Benifla 2006 Engel Class Scale Retrospective 106 78 65 Temporal only 13.5 (NA) 5.9 (NA) 24 to 162
Berkovic Engel Class Scale Prospective 135 74 NA Temporal only 11 to 58 NA 18 to 81

1995
138

Table 1. Study characteristics and participant demographics for 178 included studies (Continued)
Blount 2004 Engel Class Scale Retrospective 30 12 13 Temporal and extratem- 11.7 (4.4) NA > 30
poral
Library
Cochrane
Blume 2004 Engel Class Scale Retrospective 70 26 41 Temporal and extratem- 6 to 65 1 to 41 24 to 144
poral
Boesebeck Other scale Retrospective 81 33 51 Extratemporal only 16 to 53 2 to 43 > 24

2007
Better health.
Informed decisions.
Trusted evidence.
Boshuisen Other scale Retrospective 43 33 18 Temporal and extratem- 0 to 14 0 to 12 12 to 188
2010 poral
Brainer-Li- Engel Class Scale Retrospective 32 29 24 Temporal and extratem- 6 to 57 2 to 30 4 to 68

ma 1996 poral
Britton 1994 Engel Class Scale Retrospective 51 34 NA Temporal and extratem- 6 to 60 NA 24 to 96

poral
Caraballo Engel Class Scale Retrospective 45 33 27 Not stated 0 to 18 NA 12 to 192

2011
Cascino Engel Class Scale Retrospective 165 113 NA Temporal only 32.1 (10.5) NA > 12
1995
Chabardes Engel Class Scale Retrospective 54 47 24 Temporal only NA 2.5 to 41 48 to 100

2005
Chang 2009 Engel Class Scale Retrospective 57 45 NA Temporal and extratem- NA NA > 12
poral
Chee 1993 > 1 year seizure-free Retrospective 38 28 25 Temporal only 18 to 53 3 to 36 > 12

Chkhenkeli Engel Class Scale Retrospective 67 12 NA Temporal only NA NA >24
2007
Choi 2004a Engel Class Scale Retrospective 35 27 19 Temporal only 16 to 61 4 to 22 16 to 105
Chung 2005 Engel Class Scale Retrospective 128 58 85 Temporal and extratem- NA NA 26.9 (12)
poral
Cossu 2005 Engel Class Scale Retrospective 165 93 NA Temporal and extratem- NA NA > 12
poral
139

Cossu 2008 Engel Class Scale Retrospective 113 77 67 Temporal and extratem- 1 to 15 0 to 14 24 to 115
poral
Library
Cochrane
Costello Engel Class Scale Retrospective 42 32 21 Temporal and extratem- 45 to 66 0.5 to 55.5 13 to 173
2009 poral
Cukiert 2002 Engel Class Scale Prospective 100 89 43 Temporal only 28 (9) NA 18 to 48
Better health.
Informed decisions.
Trusted evidence.
Dagar 2011 > 1 year seizure-free Retrospective 112 89 67 Temporal and extratem- 0 to 18 0 to 15 > 12
poral
Dalmagro Engel Class Scale Retrospective 43 28 23 Not stated 22.4 (NA) 14.7 (11.5) > 12
2005
de Tisi 2011 Other scale Retrospective 615 245 287 Temporal and extratem- 16 to 63 20.7 (me- 12 to 228
poral dian)
Delbeke Engel Class Scale Retrospective 38 23 15 Temporal only 15 to 59 NA 18 to 58

1996
Dellabadia Engel Class Scale Retrospective 35 20 NA Temporal and extratem- NA NA 22 to 48

2002 poral
Devlin 2003 Engel Class Scale Retrospective 33 17 21 Temporal and extratem- 0 to 17 7.4 (medi- 12 to 96
poral an)
Donadio Engel Class Scale Retrospective 84 60 45 Temporal and extratem- 2 to 59 1 to 50 12 to 126

2011 poral
Dorward Engel Class Scale Retrospective 33 18 18 Extratemporal only 3 to 19 0.5 to 16 49.4 (NA)
2011

Duchowny > 1 year seizure-free Retrospective 31 16 16 Temporal and extratem- 0 to 3 NA > 12
1998 poral
Dunkley Engel Class Scale Retrospective 42 20 24 Temporal and extratem- 0 to 3 0 to 3 27 to 158

2011 poral
Dunlea 2010 Engel Class Scale Retrospective 199 119 95 Temporal and extratem- 1 to 61 0.5 to 43 12 to 288
poral
Elsharkawy Engel Class Scale Retrospective 218 121 129 Extratemporal only 16 to 69 1 to 65 12 to 60
2008a
140

Elsharkawy Engel Class Scale Retrospective 430 311 220 Temporal and extratem- 16 to 61 1 to 57 > 24
2009a poral
Library
Cochrane
Elsharkawy Engel Class Scale Retrospective 47 38 33 Temporal only 32 (12) 11.8 (8.8) 6 to 72
2011a
Engman Other scale Retrospective 54 35 NA Temporal only 34 (medi- NA 33.6 (me-

2004 an) dian)
Better health.
Informed decisions.
Trusted evidence.
Erba 1992 > 1 year seizure-free Retrospective 46 37 28 Temporal only 4 to 34 1 to 31 38 to 216
Erickson Engel Class Scale Retrospective 71 46 27 Temporal only 5 to 64 0 to 46 > 12

2005
Fauser 2004 Engel Class Scale Retrospective 59 35 34 Temporal only 2 to 66 NA 6 to 48
Fujiwara > 1 year seizure-free Retrospective 44 23 0 Temporal and extratem- NA NA 12 to 26

2012 poral
Garcia 1991 > 1 year seizure-free Retrospective 55 35 NA Temporal only 9 to 47 NA 12 to 48
Garcia 1994 > 1 year seizure-free Prospective 51 36 NA Temporal only NA NA 12 to 48
Gelinas 2011 Engel Class Scale Retrospective 67 52 39 Temporal and extratem- 0 to 16 2.7 (NA) 75.6 (NA)
poral
Georgakou- Engel Class Scale Retrospective 50 42 25 Temporal only 14 to 62 NA 60 to 120

lias 2008
Gilliam > 1 year seizure-free Retrospective 78 53 18 Temporal and extratem- 1 to 12 1 to 11 7 to 72

1997a poral

Gilliam > 1 year seizure-free Retrospective 33 22 40 Temporal only 9 to 50 NA > 12
1997b
Goldstein Other scale Retrospective 33 15 17 Temporal only 0 to 15 0 to 12 24 to 120

1996
Greiner 2011 > 1 year seizure-free Retrospective 54 42 NA Temporal and extratem- 0.5 to 40 7.6 (medi- > 12
poral an)
Grivas 2006 Engel Class Scale Retrospective 52 37 22 Temporal only 50 to 71 1 to 62 12 to 84

141

Gyimesi Other scale Retrospective 100 74 NA Temporal only NA NA >24
2007
Library
Cochrane
Hader 2004 Engel Class Scale Retrospective 39 21 16 Temporal and extratem- 0 to 18.5 0.19 > 18
poral
Hajek 2009 Engel Class Scale Retrospective 35 25 15 Temporal only 10 to 58 5 to 47 24 to 91
Better health.
Informed decisions.
Trusted evidence.
Hallbook > 1 year seizure-free Retrospective 110 88 71 Temporal and extratem- 0 to 18 NA 12 to 84
2010 poral
Hartley 2002 Engel Class Scale Retrospective 35 20 11 Temporal and extratem- 1 to 18 1 to 17 36 to 72

poral
Hartzfield Engel Class Scale Retrospective 56 24 35 Temporal only 8 to 70 4.5 to 41 6 to 108

2008
Hemb 2010 > 1 year seizure-free Retrospective 325 223 229 Temporal and extratem- 7.7(6.3) 4.9 (4.6) > 24
poral
Hamiwka Engel Class Scale Retrospective 38 17 NA Temporal and extratem- 0.5 to 18 NA > 24
2005 poral
Holmes > 1 year seizure-free Retrospective 44 22 17 Temporal only 14 to 55 3 to 46 12 to 48

1997
Holmes Other scale Retrospective 126 54 71 Temporal and extratem- 6 to 69 1 to 46 24 to 72

2000 poral
Jack 1992 > 1 year seizure-free Retrospective 50 34 27 Temporal only 14 to 51 2 to 37 12 to 34
Jaramil- Engel Class Scale Retrospective 67 43 52 Temporal only 27 (11) 20 (10) > 24

lo-Betancur
2009
Janszky Engel Class Scale Retrospective 84 60 NA Temporal only NA NA > 24

2003a
Janszky Engel Class Scale Retrospective 147 123 NA Temporal only NA NA >6
2003b
Jayakar Engel Class Scale Retrospective 101 44 60 Temporal and extratem- 1.5 to 21 NA > 24
2008 poral
142

Jayalakshmi Engel Class Scale Retrospective 78 50 44 Temporal and extratem- 2 to 16 1 to 16 12 to 58
2011 poral
Library
Cochrane
Jeha 2006 > 1 year seizure-free Retrospective 371 231 NA Temporal only NA NA > 12
Jehi 2012 > 1 year seizure-free Retrospective 312 165 149 Temporal only 2.5 to 74 1 to 64 42 (20.4)
Jennum > 1 year seizure-free Retrospective 64 42 45 Temporal and extratem- 8 to 52 1 to 38 > 12
Better health.
Informed decisions.
Trusted evidence.
1993 poral
Jeong 1999 Other scale Not stated 93 78 54 Temporal only 9 to 51 NA 18 to 33
Kan 2008 Engel Class Scale Retrospective 58 43 34 Temporal and extratem- 2 to 21 0.5 to 15 12 to 96
poral
Kang 2009 Engel Class Scale Retrospective 244 194 108 Temporal only 18 to 68 NA 12 to 204
Kanner 2009 Engel Class Scale Retrospective 100 86 60 Temporal only 31.2 (10.7) 12.7 (NA) 24 to 168
Kilpatrick Engel Class Scale Prospective 50 39 28 Temporal only 16 to 57 NA 12 to 38

1997
Kim 2009 > 1 year seizure-free Retrospective 166 94 102 Temporal and extratem- 3 to 51 0.5 to 37 95.3 (NA)
poral
Kim 2010a Engel Class Scale Retrospective 177 75 121 Temporal and extratem- 11 to 51 1 to 50 24 to 180
poral
Kim 2010b > 1 year seizure-free Retrospective 40 14 26 Temporal and extratem- 4 to 51 1 to 48 > 24
poral
Kloss 2002 Engel Class Scale Combination 68 34 30 Temporal and extratem- 0 to 16 NA 12 to 108

poral
Knowlton Engel Class Scale Prospective 62 37 33 Temporal and extratem- 1 to 60 NA > 12

2008 poral
Kral 2007 Other scale Retrospective 40 27 23 Temporal and extratem- 5 to 47 1 to 45 97 (54)

poral
Krsek 2013 Engel Class Scale Retrospective 106 64 49 Not stated 0 to 30 NA > 24
143

Kuzniecky Other scale Prospective 34 23 14 Temporal only 7 to 38 >2 12 to 30
1993
Library
Cochrane
Kwan 2010 Engel Class Scale Retrospective 41 27 22 Extratemporal only 0 to 17.5 0 to 16 72 (NA)
Lackmayer Other scale Retrospective 45 32 21 Temporal only 40.8 (10.2) 13.0 (10.1) > 12
2013
Better health.
Informed decisions.
Trusted evidence.
Lee 2006 Other scale Retrospective 51 37 29 Temporal only 16 to 50 4 to 38 > 24
Lee 2008 Engel Class Scale Not stated 71 38 44 Extratemporal only 12 to 57 3 to 34 > 24
Lee 2010 Engel Class Scale Retrospective 52 36 31 Temporal only 9 to 54 1 to 33 14 to 42
Lee 2011 Engel Class Scale Retrospective 40 37 23 Temporal only 1 to 15 0 to 14 11 to 151
Lei 2008 > 1 year seizure-free Retrospective 196 180 210 Temporal and extratem- 17 to 66 NA > 48
poral
Li 1997 Engel Class Scale Retrospective 51 39 23 Temporal and extratem- 8 to 73 NA 12 to 157

poral
Li 1999 Engel Class Scale Not stated 38 16 23 Temporal and extratem- 14 to 63 NA 12 to 180
poral
Liang 2010 Engel Class Scale Prospective 60 40 34 Temporal and extratem- 16.7 (NA) NA > 24
poral
Liang 2012 Engel Class Scale Retrospective 206 173 94 Temporal and extratem- 6 to 14 2 to 14 > 12
poral
Liava 2012 Engel Class Scale Retrospective 52 39 NA Extratemporal only 1 to 26 7.9 (NA) 18 to 162

Lopez-Gon- > 1 year seizure-free Retrospective 86 65 63 Temporal only 1 to 18 1 to 17 > 12
zalez 2012
Lorenzo Engel Class Scale Retrospective 48 17 NA Extratemporal only NA NA > 12

1995
Madhavan Engel Class Scale Retrospective 70 37 39 Temporal and extratem- 9.9 (10.2) NA
2007 poral
144

Mani 2006 Engel Class Scale Retrospective 122 86 72 Temporal and extratem- 8.2 (5.5) 5.2 (4.6) >6
poral
Library
Cochrane
Mathern Other scale Retrospective 198 78 111 Temporal and extratem- 0 to 37 0 to 31 6 to 120
1999 poral
McIntosh Other scale Retrospective 81 11 38 Extratemporal only 4 to 60 16 (medi- 12 to 212

2012 an)
Better health.
Informed decisions.
Trusted evidence.
Mihara 2004 Engel Class Scale Retrospective 357 269 NA Temporal and extratem- 2 to 55 1 to 40 24 to 196
poral
Miserocchi Engel Class Scale Retrospective 68 58 43 Temporal only 1 to 15 0.5 to 14 > 12

2013
Morino 2009 > 1 year seizure-free Retrospective 62 55 24 Temporal only 34.4 (NA) NA > 12
Morris 1998 Engel Class Scale Retrospective 36 26 22 Temporal and extratem- 2 to 56 1 to 29 6 to 41

poral
O'Brien > 1 year seizure-free Retrospective 46 36 24 Temporal only 16 to 58 2 to 43 12 to 33

1996
O'Brien Engel Class Scale Retrospective 36 14 23 Temporal and extratem- 1 to 56 NA 12 to 40

2000 poral
Oertel 2005 > 1 year seizure-free Prospective 35 20 11 Temporal only 35.1 (NA) NA 12 to 42
Paglioli Engel Class Scale Prospective 160 143 88 Temporal only 8 to 62 3 to 60 24 to 132
2006
Paolicchi > 1 year seizure-free Retrospective 75 44 40 Temporal and extratem- 0 to 12 NA 12 to 120

2000 poral
Park 2002 Engel Class Scale Retrospective 148 108 76 Temporal and extratem- 0.5 to 18 0 to 19 0 to 166
poral
Park 2006 Engel Class Scale Retrospective 30 20 19 Temporal and extratem- 0 to 13 1 to 18 12 to 64

poral
Perego 2009 Engel Class Scale Retrospective 37 29 19 Temporal only 33 (10) 19 (NA) > 36
145

Perry 2010 Engel Class Scale Retrospective 83 34 50 Temporal and extratem- NA 0 to 17 > 24
poral
Library
Cochrane
Phi 2009 Engel Class Scale Retrospective 87 80 49 Temporal only 1 to 62 0 to 48 12 to 128
Phi 2010 Other scale Retrospective 41 20 20 Temporal and extratem- 1 to 17 1 to 11 24 to 153

poral
Better health.
Informed decisions.
Trusted evidence.
Pin- Engel Class Scale Retrospective 70 33 27 Extratemporal only 1 to 52 1 to 36 59.1 (30.5)
heiro-Mar-
tins 2012
Prevedello Engel Class Scale Retrospective 84 53 48 Temporal only 19 to 43 10 to 33 15 to 44

2000
Raabe 2012 Engel Class Scale Retrospective 76 57 55 Not stated 6 to 67 0 to 53 24 to 200
Radhakrish- Engel Class Scale Combination 175 134 77 Temporal only 7 to 86 0 to 81 24 to 68

nan 1998
Rausch 2003 > 1 year seizure-free Not stated 42 21 21 Temporal only NA NA 109 to 228
Remi 2011 Engel Class Scale Retrospective 154 138 NA Temporal and extratem- NA NA 22 to 228
poral
Roberti 2007 Engel Class Scale Retrospective 42 27 17 Temporal only 10 to 52 3 to 51 44 to 121
Rossi 1994 Other scale Retrospective 138 86 100 Temporal and extratem- 1 to 46 >3 > 36
poral
Russo 2003 Engel Class Scale Retrospective 101 64 67 Temporal and extratem- 0 to 53 0 to 46 > 12
poral

Sagher 2012 Engel Class Scale Retrospective 96 85 46 Temporal only 17 to 59 1 to 57 > 12
Sakamoto Other scale Retrospective 36 19 16 Temporal only 12 to 58 4 to 49 24 to 63

2009
Salanova Engel Class Scale Retrospective 89 53 46 Temporal only 8 to 53 1 to 43 12 to 96

1994
Sarkis 2012 Engel Class Scale Retrospective 62 38 34 Temporal only 1 to 56 NA 6 to 170

146

Schramm Engel Class Scale Prospective 207 152 100 Temporal only 39.7 (13.2) 22 (medi- > 12
2011 an)
Library
Cochrane
Seymour Engel Class Scale Retrospective 291 153 159 Temporal only 3 to 59 NA 28 to 144
2012
Sinclair Engel Class Scale Retrospective 77 57 39 Temporal and extratem- 0 to 16 0 to 15 12 to 144

2003 poral
Better health.
Informed decisions.
Trusted evidence.
Sindou 2006 Engel Class Scale Not stated 100 85 42 Temporal only 18 to 58 NA 12 to 120
Sola 2005 Engel Class Scale Retrospective 137 74 63 Temporal only 12 to 69 1 to 44 24 to 138
Spencer > 1 year seizure-free Prospective 355 264 174 Temporal and extratem- NA NA > 12
2005 poral
Sperling Engel Class Scale Prospective 51 41 31 Temporal only 17 to 59 NA 21 to 64

1992
Stavrou Engel Class Scale Retrospective 53 45 31 Temporal and extratem- NA 3.4 (0.8) >24
2008 poral
Suppiah Engel Class Scale Retrospective 174 94 75 Temporal only 10 to 61 NA 57 (32)

2009
Swartz 1992 > 1 year seizure-free Prospective 34 27 NA Temporal only 29 (NA) NA 20 to 71
Tanriverdi Engel Class Scale Retrospective 256 160 123 Temporal only 30.3 (10.5) 20.6 (10.7) > 12
2010
Tatum 2008 > 1 year seizure-free Retrospective 39 22 15 Temporal only 33.8 (10.9) 14.5 (NA) 30.8

Terra-Bus- Engel Class Scale Prospective 107 62 55 Temporal and extratem- 10.2 (5.4) 6.4 (4.8) 12 to 108
tamante poral
2005a
Terra-Bus- Engel Class Scale Prospective 35 27 11 Temporal only 1 to 18 0 to 15 12 to 84

tamante
2005b
Tezer 2008 Engel Class Scale Retrospective 109 90 42 Temporal only 15 to 52 18.5 (7.8) 57.4 (30.6)
147

Theodore > 1 year seizure-free Prospective 41 26 19 Temporal only 34.2 (9.5) 25 (12) 12 to 132
2012
Library
Cochrane
Tigaran Engel Class Scale Retrospective 65 32 41 Extratemporal only 4 to 50 1 to 40 12 to 144
2003
Tripathi Engel Class Scale Retrospective 55 29 33 Temporal and extratem- 1 to 45 NA 60 to 120

2008 poral
Better health.
Informed decisions.
Trusted evidence.
Trottier Engel Class Scale Retrospective 96 73 69 Extratemporal only 0.5 to 52 16 (NA) > 12
2008
Urbach 2007 Engel Class Scale Retrospective 42 25 20 Extratemporal only 7 to 50 2 to 42 > 12
Ure 2009 Engel Class Scale Retrospective 74 28 0 Not stated 14 to 53 NA 12 to 336
Velasco Engel Class Scale Prospective 163 93 NA Not stated NA NA 14 to 73

2011
Walz 2003 > 1 year seizure-free Not stated 98 85 44 Temporal only 36.2 (10.7) 25.2 (10.5) 12 to 90
Weinand > 1 year seizure-free Retrospective 89 57 NA Temporal only 6 to 52 2 to 42 12 to 54

1992
Wellmer Engel Class Scale Retrospective 165 106 NA Not stated NA NA 24 to 103
2012
Wid- > 1 year seizure-free Retrospective 48 22 26 Temporal and extratem- 2 to 56 0 to 27 32.4 (me-
dess-Walsh poral dian)
2007
Wiebe 2001 > 1 year seizure-free Prospective 36 23 NA Temporal only NA NA > 12

Wieshmann Engel Class Scale Retrospective 76 46 32 Not stated NA NA > 12
2008
Wray 2012 Engel Class Scale Retrospective 52 36 24 Temporal and extratem- 0 to 17 NA > 12
poral
Wyler 1995 > 1 year seizure-free Prospective 70 38 33 Temporal only NA NA > 12
Wyllie 1998 Engel Class Scale Not stated 136 92 78 Temporal and extratem- 13 to 20 NA 12 to 88
poral
148

Yang 2011 Other scale Retrospective 99 27 63 Not stated 8 to 59 NA 12 to 72
Library
Cochrane
Yeon 2009 Engel Class Scale Retrospective 60 50 37 Temporal and extratem- NA 1 to 26 34.9 (22.5)
poral
Yu 2009 Engel Class Scale Retrospective 43 26 28 Extratemporal only 4 to 43 1 to 30 33.6 (16.8)
Yu 2012a > 1 year seizure-free Retrospective 100 61 NA Temporal and extratem- 26.8 (7.6) 14.3 (8.3) 12 to 108
Better health.
Informed decisions.
Trusted evidence.
poral
Yu 2012b > 1 year seizure-free Retrospective 222 148 NA Temporal and extratem- 12.5 (8.1) 6.6 (4.3) 12 to 108
poral
Zangaladze > 1 year seizure-free Combination 99 66 NA Temporal and extratem- NA NA > 24

2008 poral
Zentner Engel Class Scale Retrospective 167 103 82 Temporal only 3 to 64 2 to 52 12 to 72

1995
Zentner Engel Class Scale Prospective 56 30 39 Extratemporal only 1 to 49 2 to 35 20 to 85

1996
1Age at surgery (years), duration of epilepsy (years) and follow-up (months) expressed as minimum and maximum years or months where reported. If minimum and maximum
not available reported as mean (SD) or median (noted in table). Follow-up expressed as minimum only (e.g. > 12 months) for 58 studies.
NA: not available

Table 2. Quality assessment of included studies according to EPHPP tool
Author Design Study design Selection Study Con- Blinding Data col- Withdrawals Global
bias rat- design founders rating lection rating rating

ing rating rating rating
Aaberg 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Adam 1996 Prospective Cohort Moderate Moder- Strong Weak Weak Strong Weak
ate
Adelson 1992 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
149

Table 2. Quality assessment of included studies according to EPHPP tool (Continued)
Alfstad 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Library
Cochrane
Althausen Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
2013 ate
Arruda 1996 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Better health.
Informed decisions.
Trusted evidence.
Awad 1991 Not stated Cohort Moderate Moder- Strong Weak Strong Weak Weak
ate
Babini 2013 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Battaglia 2006 Not stated Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate
Baumann Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
2007 ate
Bautista 2003 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Bell 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Benifla 2006 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Berkovic 1995 Prospective Cohort Moderate Moder- Strong Weak Weak Strong Weak
ate

Blount 2004 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Blume 2004 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Boesebeck Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
2007 ate
150

Boshuisen Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2010 ate
Library
Cochrane
Brainer-Lima Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
1996 ate
Britton 1994 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Better health.
Informed decisions.
Trusted evidence.
Caraballo Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
2011 ate
Cascino 1995 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
Chabardes Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2005 ate
Chang 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Chee 1993 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
Chkhenkeli Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2007 ate
Choi 2004a Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Chung 2005 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate

Cossu 2005 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Cossu 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Costello 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
151

Cukiert 2002 Prospective Cohort Moderate Moder- Strong Weak Weak Strong Weak
ate
Library
Cochrane
Dagar 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Dalmagro Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2005 ate
Better health.
Informed decisions.
Trusted evidence.
de Tisi 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Delbeke 1996 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Dellabadia Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2002 ate
Devlin 2003 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Donadio 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Dorward 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Duchowny Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
1998 ate
Dunkley 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate

Dunlea 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Elsharkawy Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2008a ate
2009a ate
152

2011a ate
Library
Cochrane
Engman 2004 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Weak
ate
Erba 1992 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Better health.
Informed decisions.
Trusted evidence.
Erickson 2005 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Fauser 2004 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
Fujiwara 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Garcia 1991 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Garcia 1994 Prospective Cohort Moderate Moder- Strong Weak Strong Weak Weak
ate
Gelinas 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Georgakoulias Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2008 ate
Gilliam 1997a Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate

Gilliam 1997b Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Goldstein Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
1996 ate
Greiner 2011 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
153

Grivas 2006 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Library
Cochrane
Gyimesi 2007 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Hader 2004 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Better health.
Informed decisions.
Trusted evidence.
Hajek 2009 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
Hallbook 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Hartley 2002 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Hartzfield Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2008 ate
Hemb 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Hamiwka Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2005 ate
Holmes 1997 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
Holmes 2000 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate

Jack 1992 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Janszky Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
2003a ate
Janszky Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2003b ate
154

Jaramillo-Be- Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
tancur 2009 ate
Library
Cochrane
Jayakar 2008 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Jayalakshmi Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
2011 ate
Better health.
Informed decisions.
Trusted evidence.
Jeha 2006 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Jehi 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Jennum 1993 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Jeong 1999 Not stated Cohort Moderate Moder- Strong Weak Strong Weak Weak
ate
Kan 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Kang 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Kanner 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Kilpatrick Prospective Cohort Moderate Moder- Strong Weak Weak Weak Weak
1997 ate

Kim 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Kim 2010a Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Kim 2010b Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
155

Kloss 2002 Combination Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate
Library
Cochrane
Knowlton Prospective Cohort Moderate Moder- Strong Weak Strong Weak Weak
2008 ate
Kral 2007 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Better health.
Informed decisions.
Trusted evidence.
Krsek 2013 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Kuzniecky Prospective Cohort Moderate Moder- Strong Weak Weak Strong Weak
1993 ate
Kwan 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Lackmayer Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Weak
2013 ate
Lee 2006 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Lee 2008 Not stated Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate
Lee 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Lee 2011 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate

Lei 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Li 1997 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Li 1999 Not stated Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate
Liang 2010 Prospective Randomised controlled trial Moderate Strong Strong Moderate Strong Strong Strong
156

Liang 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Library
Cochrane
Liava 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Lopez-Gonza- Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
lez 2012 ate
Better health.
Informed decisions.
Trusted evidence.
Lorenzo 1995 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Madhavan Retrospective Retrospective case series Strong Moder- Strong Weak Weak
2007 ate
Mani 2006 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Mathern 1999 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
McIntosh Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2012 ate
Mihara 2004 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Miserocchi Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
2013 ate
Morino 2009 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate

Morris 1998 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
O'Brien 1996 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
O'Brien 2000 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Oertel 2005 Prospective Cohort analytic Moderate Strong Moderate Weak Weak Weak Weak
157

Paglioli 2006 Prospective Cohort Moderate Moder- Strong Weak Strong Strong Moderate
ate
Library
Cochrane
Paolicchi 2000 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Park 2002 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Better health.
Informed decisions.
Trusted evidence.
Park 2006 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Perego 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Perry 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Phi 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Phi 2010 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Pinheiro-Mar- Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
tins 2012 ate
Prevedello Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
2000 ate
Raabe 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate

Radhakrish- Combination Cohort Moderate Moder- Strong Weak Strong Weak Weak
nan 1998 ate
Rausch 2003 Not stated Cohort Moderate Moder- Strong Weak Strong Weak Weak
ate
Remi 2011 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
158

Roberti 2007 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Library
Cochrane
Rossi 1994 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Russo 2003 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Better health.
Informed decisions.
Trusted evidence.
Sagher 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Sakamoto Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2009 ate
Salanova Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
1994 ate
Sarkis 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Schramm Prospective Randomised controlled trial Moderate Strong Strong Strong Weak Strong Moderate
2011
Seymour 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Sinclair 2003 Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
ate
Sindou 2006 Not stated Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate

Sola 2005 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Spencer 2005 Prospective Cohort Moderate Moder- Strong Weak Strong Weak Weak
ate
Sperling 1992 Prospective Cohort Moderate Moder- Strong Weak Strong Strong Moderate
ate
159

Stavrou 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Library
Cochrane
Suppiah 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Swartz 1992 Prospective Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate
Better health.
Informed decisions.
Trusted evidence.
Tanriverdi Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
2010 ate
Tatum 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Terra-Busta- Prospective Cohort Moderate Moder- Strong Weak Strong Weak Weak
mante 2005a ate
Terra-Busta- Prospective Cohort Moderate Moder- Strong Weak Strong Weak Weak
mante 2005b ate
Tezer 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Theodore Prospective Cohort Moderate Moder- Strong Weak Weak Weak Weak
2012 ate
Tigaran 2003 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Tripathi 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate

Trottier 2008 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Urbach 2007 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Ure 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Velasco 2011 Prospective Cohort analytic Strong Strong Strong Weak Weak Strong Weak
160

Walz 2003 Not stated Cohort Moderate Moder- Strong Weak Weak Weak Weak
ate
Library
Cochrane
Weinand 1992 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Wellmer 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Better health.
Informed decisions.
Trusted evidence.
Wid- Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
dess-Walsh ate
2007
Wiebe 2001 Prospective Randomised controlled trial Strong Strong Strong Weak Strong Strong Moderate
Wieshmann Retrospective Retrospective case series Moderate Moder- Strong Weak Weak Not applicable Weak
2008 ate
Wray 2012 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate
Wyler 1995 Prospective Randomised controlled trial Moderate Strong Strong Strong Strong Strong Strong
Wyllie 1998 Not stated Cohort Strong Moder- Strong Weak Strong Weak Weak
ate
Yang 2011 Retrospective Retrospective case series Moderate Moder- Strong Weak Strong Not applicable Moderate
ate
Yeon 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Strong Not applicable Moderate
ate

Yu 2009 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Yu 2012a Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
Zangaladze Combination Cohort Moderate Moder- Strong Weak Weak Weak Weak
2008 ate
Zentner 1995 Retrospective Retrospective case series Strong Moder- Strong Weak Weak Not applicable Weak
ate
161

Informed decisions.

Weak
Strong
Weak
Weak
Strong
Moder-
ate
Strong
Cohort
Prospective
Zentner 1996

Informed decisions.


Table 3. Summary of EPHPP tool quality assessment component ratings
Quality assessment criteria Number and proportion of studies
Method of identifying participants Retrospective 143 (81%)
Prospective 21 (12%)
Combination 3 (2%)
Not stated 9 (5%)
A. Selection bias
(Q1) Are the individuals selected to participate in the study likely to Very likely 124 (70%)
be representative of the target population? Somewhat likely 52 (30%)
(Q2) What percentage of selected individuals agreed to participate? 89% to 100% 6 (4%)
Can't tell 27 (15%)
Not applicable 143 (81%)
Selection bias rating Moderate 62 (35%)
Strong 114 (65%)
B. Study design Retrospective case series 143 (81%)
Cohort 27 (15%)
Cohort analytic 2 (1%)
Randomised controlled 4 (3%)

trial
Study design rating Moderate 170 (97%)
Strong 6 (3%)
C. Confounders
(Q1) Were there important differences between groups prior to the interven- No 175 (99%)
tion?
Unclear/ Can't tell 1 (1%)
(Q2) If yes, indicate the percentage of relevant confounders that were con- Unclear/ Can't tell 1 (1%)
trolled
(either in the design (e.g. stratification, matching) or analysis)?
Confounders rating Strong 175 (99%)
Moderate 1 (1%)
D. Blinding

Informed decisions.

Table 3. Summary of EPHPP tool quality assessment component ratings (Continued)

(Q1) Was (were) the outcome assessor(s) aware of the intervention Yes 173 (98%)
or exposure status of participants? No 2 (1%)
Can't tell 1 (1%)
(Q2) Were the study participants aware of the research question? Yes 173 (98%)
No 2 (1%)
Can't tell 1 (1%)
Blinding rating Strong 2 (1%)
Moderate 1 (1%)
Weak 173 (98%)
E. Data collection methods
(Q1) Were data collection tools shown to be valid? Yes 90 (51%)
Unclear/can't tell 86 (49%)
(Q2) Were data collection tools shown to be reliable? Yes 90 (51%)
Unclear/can't tell 86 (49%)
Data collection rating Strong 90 (51%)
Weak 86 (49%)
F. Withdrawals and drop-outs
(Q1) Were withdrawals and drop-outs reported in terms of numbers Yes 12 (7%)
and/or reasons per group? No 21 (12%)
(Q2) Indicate the percentage of participants completing the study. 80% to 100% 12 (7%)
Can't tell 21 (12%)
Withdrawals rating Strong 12 (7%)
Weak 21 (12%)
G. Intervention integrity
(Q1) What percentage of participants received the allocated 80% to 100% 176 (100%)

Informed decisions.

Table 3. Summary of EPHPP tool quality assessment component ratings (Continued)

intervention or exposure of interest?
(Q2) Was the consistency of the intervention measured? Yes 158 (90%)
No 18 (10%)
(Q3) Is it likely that subjects received an unintended intervention (contamina- No 176 (100%)
tion
or co-intervention) that may influence the results?
H. Analyses
(Q1) Unit of allocation Individual 176 (100%)
(Q2) Unit of analysis Individual 176 (100%)
(Q3) Are the statistical methods appropriate for the study design? Yes 153 (87%)
No stats 23 (13%)
(Q4) Is the analysis performed by intervention allocation status Yes 176 (100%)
(i.e. intention-to-treat) rather than the actual intervention received?
Global rating Strong 2 (1%)
Moderate 78 (44%)
Weak 96 (55%)

Table 4. Risk of bias assessment of four randomised controlled trials
Study Random Allocation conceal- Blinding of partici- Incomplete out- Selective reporting Other
sequence ment (selection bias) pants, personnel and come data (attri- (reporting bias) bias
generation outcome assessors tion bias)
(selection (performance and de-
bias) tection bias)
Liang High Unclear Unclear Low Low Low

2012:
judge-
ment
Sup- Quasi-ran- No information provid- No information provid- All participants Seizure outcome, No
port domisation ed ed followed for 2 complications and be- other
for based on years and includ- havioural changes de- bias
judge- odd or even ed in the outcome scribed in the meth- de-
ment patient ID at 2 years ods section well re- tected
number, ported in the results.
not an ad- No protocol available
equate to assess planned out-
method of comes a priori
randomisa-
tion

Informed decisions.

Table 4. Risk of bias assessment of four randomised controlled trials (Continued)

SchrammLow Low Low Low Low Low
2011:
judge-
ment
Sup- A comput- Each centre received Neither the study par- No losses to fol- Seizure outcome, ex- No
port er-generat- the randomisation ticipants nor the epilep- low-up, all partic- tent of resection and other
for ed blocked codes in numbered and tologists assessing the ipants included complications de- bias
judge- randomisa- sealed envelopes, each outcomes were told of in analysis of out- scribed in the meth- de-
ment tion list was envelope allocating the the result of the ran- come at 1 year ods section well re- tected
prepared extent of resection for domisation, i.e. they ported in the results.
one particular patient were not aware whether No protocol available
in that centre. The en- a short or a long resec- to assess planned out-
velopes were to be used tion had been done; the comes a priori
in the given sequence only person knowing
for the patients as they this was the surgeon.
were included into the OR notes did not men-
study. They had to be tion the planned resec-
opened in the OR not tion extent. In partic-
before the morning of ular, the persons per-
surgery. Thus, the type forming the MRI vol-
of surgery (selective umetry were blinded to
amygdalohippocam- group assignment
pectomy or lobectomy)
was determined be-
fore the envelope was
opened, but the con-
tent of the envelope de-
termined the mesial ex-
tent of resection
Wiebe Unclear Low Unclear Low Low Low

2001:
judge-
ment
Sup- Partic- The random assign- Blinding was not possi- Full details of at- Seizure outcome No
port ipants ments were prepared ble of anyone involved trition given: "No and quality of life de- other
for were "ran- outside the study cen- (surgical versus med- patients were scribed in the meth- bias
judge- domly as- tre and delivered in ical treatment); unclear lost to follow-up. ods section well re- de-
ment signed…af- sealed, opaque, se- if outcome was influ- There were no ported in the results. tected
ter stratifi- quentially numbered enced by this crossovers from No protocol available
cation ac- envelopes the medical group to assess planned out-
cording to to the surgical comes a priori
the pres- group. One pa-
ence or ab- tient in the med-
sence of ical group died
general- (a sudden, unex-
ized motor plained death) 7.5
seizures"; months into the
no further study. No deaths
informa- occurred in the
tion given surgical group."
Intention-to-treat
approach taken
Wyler Unclear Unclear Low Low Low Low

1995:

Informed decisions.

Table 4. Risk of bias assessment of four randomised controlled trials (Continued)

judge-
ment
Sup- Study de- No information provid- Participants and out- All participants in- Seizure outcome and No
port scribed ed come assessors blind- cluded in assess- neuropsychological other
for as ran- ed, not possible to blind ment of outcome outcomes described bias
judge- domised, surgeons by design at 1 year, inten- in the methods sec- de-
ment no further tion-to-treat ap- tion well reported in tected
informa- proach taken the results. No proto-
tion given col available to assess
planned outcomes a
priori

OR: operating room


Table 5. Risk of bias in prognostic studies according to QUIPS tool
Author 1. Study participation 2. Study 3. Prognostic factor measure- 4. Outcome measure- 5. Study con- 6. Statistical
attrition ment ment founding analysis and re-
Library
Cochrane
porting
Judgement Risk of Judge-

Risk of Judgement Risk of Judgement Risk of Judge- Risk of Judgement Risk of
bias ment
bias bias bias ment bias bias
Al- Population of inter- mod- Ret-NA Most variables well de- mod- Outcome is high Most mod- Insufficient infor- high
Better health.
Informed decisions.
Trusted evidence.
thausen est well described erate ro- fined, some not. Un- erate patient-re- vari- erate mation provid-
2013 and sample re- spec- clear how data were ported there- ables ed in the meth-
cruited seems to tive collected and whether fore likely to well ods, some vari-
match this, how- so data were complete be highly sub- de- ables included in
ever sample re- NA jective, al- fined, the model with-
stricted to those so unclear some out prespecifica-
who returned post- when the out- not. tion and no defin-
operative ques- come was Un- itions, results se-
tionnaires, may measured clear lectively reported
exclude some of and if addi- how
those with poorer tional med- data
intellectual/seizure ical records were
outcomes? were used for col-
some/all pa- lect-
tients ed and
whether
data
were
com-
plete.
Some
adjust-
ment
seems
to have

been
done
but not
for all
rele-
vant
vari-
ables
Boese- Population of in- mod- Ret-NA Prognostic factors en- mod- Difficult out- mod- NA for NA Model is well mod-
beck terest is well de- erate ro- tering the model are erate come to mea- erate stud- described and erate
168
2007 scribed and sam- spec- well described but not sure in a valid ies of a based on clinical

Table 5. Risk of bias in prognostic studies according to QUIPS tool (Continued)
ple seems to match tive much information giv- and reliable single value rather than
this but unclear on so en on how these data way, but out- group statistical signif-
Library
Cochrane
the recruitment NA were collected come defin- design icance however
methods and inclu- ition is clear not all P values
sion criteria and the same are reported and
for all pa- regression coeffi-
tients but cients are report-
method of ed rather than
collecting meaningful effect
Better health.
Informed decisions.
Trusted evidence.
outcome da- sizes
ta is not spec-
ified
Cossu Population of inter- low Ret-NA All variables well de- low Difficult out- mod- Age at mod- Analysis is de- high
2008 est well described, ro- scribed, measurement come to mea- erate surgery erate scribed clearly
sample well de- spec- techniques specified sure in a valid and but variables in
scribed, sample is tive and patients all came and reliable pathol- the model are
representative of so from a single centre way; mea- ogy based on statis-
the source popula- NA so likely followed the sured with a seem tical significance
tion same protocol known scale, to have and results are
last follow-up been presented only
means vari- consid- for statistically
ability across ered significant vari-
measurement as con- ables
time founders,
un-
clear
what
are
con-
founders
and
what

are
prog-
nostic
factors
Cossu Population of inter- mod- Ret-NA Variables are well de- mod- Difficult out- mod- NA for NA Methods are well mod-
2005 est well described erate ro- fined but very little in- erate come to mea- erate stud- described on the erate
and original sam- spec- formation on how the sure in a valid ies of a model and vari-
ple matches this tive data were collected and reliable single ables but report-
population, howev- so and analysed way; informa- group ing of results is
er not all original NA tion provided design highly selective
sample are includ- on outcome
169

ed in final analyses classification with no useful in-
and characteristics but not when formation at all
Library
Cochrane
not described for outcome was
the subgroup of in- measured
terest who did un- and how out-
dergo surgery come was
measured
ElsharkawyPopulation of in- low Ret-NA All variables well de- mod- Difficult vari- mod- De- mod- Model is appro- high
Better health.
Informed decisions.
Trusted evidence.
2008a terest and sample ro- fined, pre-operative erate able to mea- erate spite erate priate for the
are well described, spec- workup seems the sure in a valid all re- data but vari-
seem to match tive same but data collect- and reliable sults ables are includ-
so ed retrospectively in a way; outcome being ed based on sta-
NA variety of ways which well defined pre- tistical signifi-
could have led to vari- overall but sent- cance and results
ability across patients unclear at ed ac- are selectively re-
which time cord- ported
point model- ing to
ling was con- "Group",
ducted and differ-
potential vari- ences
ability in da- be-
ta collection tween
across pa- these
tients groups
don't
seem
to be
consid-
ered
and
un-
clear

if any
vari-
ables
are
being
treated
as con-
founders
or as
prog-
nostic
170

vari-
ables
Library
Cochrane
ElsharkawyPopulation of inter- low Ret-NA Most variables well mod- Difficult out- low NA for NA Model is appro- high
2009a est is well defined ro- defined, data record- erate come to mea- stud- priate for the
and sample recruit- spec- ed and collected fol- sure in a valid ies of a data but vari-
ed seems to match tive lowing the same pro- and reliable single ables are includ-
this. Very clear how so tocol, however defi- way but the group ed based on sta-
sample was recruit- NA nitions of some vari- methods design tistical signifi-
Better health.
Informed decisions.
Trusted evidence.
ed, large sample ables unclear due to seem satis- cance; some vari-
size variable terminology factory (same able definitions
which makes it unclear time points, unclear and re-
which variables have single proto- sults are selec-
been analysed col, multiple tively reported
data sources)
Gelinas Population of in- mod- Ret-NA All variables well de- mod- Difficult out- low Po- mod- Results well re- mod-
2011 terest and sample erate ro- fined and method of erate come to mea- tential erate ported but un- erate
recruited well de- spec- measurement seems sure in a valid "con- clear exactly how
scribed, seem com- tive valid and consistent and reliable founder" many variables
patible. Only un- so for most. Some vari- way. Mea- Type of went into the
certainty around NA ability possible for one sured at a sin- surgery model
the definition of of the most important gle time point well
paediatric variables in the study, for analysis de-
subjective choice and judged fined
by an expert but
so seems rea- choice
sonable was
sub-
jective
and
likely
to vary

across
pa-
tients.
Differ-
ences
be-
tween
groups
consid-
ered
and
this
171

vari-
able
Library
Cochrane
includ-
ed in
regres-
sion
analy-
sis
Better health.
Informed decisions.
Trusted evidence.
Jan- Population of inter- high Ret-NA Prognostic factors well mod- Difficult out- mod- All ap- mod- Methods describe high
szky est well described, ro- defined including defi- erate come to mea- erate parent erate models quite
2003a however given the spec- nitions of key variables sure in a valid rele- well; unclear how
clinical question of tive (IEDS, CFCs), unclear if and reliable vant it was decided
interest, the sam- so patients were submit- way. Outcome con- which variables
ple seems quite se- NA ted to one or two cen- recorded ac- founders went into the
lective tres so unclear if data cording to a exam- prognostic mod-
collection methods are known scale ined, el and results are
the same but unclear well selectively re-
when out- de- ported
come was fined
measured and
(possibly vari- ad-
able across justed
patients) and for in
little detail analy-
given about sis of
how data prog-
were collect- nos-
ed tic fac-
tors,
but un-
clear
if pa-

tients
were
sub-
mit-
ted to
1 or 2
centres
so un-
clear
if da-
ta col-
lection
172
meth-

ods
are the
Library
Cochrane
same
Jen- Population of in- mod- Ret-NA Prognostic factors en- mod- Difficult out- low NA for NA Very little infor- high
num terest is well de- erate ro- tering the model are erate come to mea- stud- mation given
1993 scribed and sam- spec- mostly well described sure in a valid ies of a about the model
ple seems to match tive (except for extent of and reliable single and only signifi-
this but unclear on so resection) but not way, howev- group cant P values re-
Better health.
Informed decisions.
Trusted evidence.
the recruitment NA much information giv- er outcome design ported
methods and inclu- en on how these data data are col-
sion criteria were collected lected in the
same way for
all patients
and measured
at the same
time
Kim Population of inter- low Ret-NA Definitions and mea- mod- Difficult out- high NA for NA Methods well de- high
2009 est well described, ro- surements of some erate come to mea- stud- scribed but vari-
sample well de- spec- factors unclear sure in a valid ies of a ables entered the
scribed, seem to tive and reliable single model based on
match so way; not de- group statistical signif-
NA fined accord- design icance and re-
ing to a scale sults selectively
such as En- reported
gel Class, no
information
on when out-
come is mea-
sured
Kim Population of in- mod- Ret-NA Most variables are well mod- Seizure out- mod- NA for NA Unclear which mod-

2010a terest and sam- erate ro- described but are tech- erate come defined erate stud- variables went in- erate
ple well defined spec- nical, not clear if pa- but not the ies of a to the model and
but unclear if pa- tive tients came from a sin- time point single results selectively
tients were recruit- so gle centre so different measured and group reported
ed from a single NA protocols may have unclear how design
centre or multiple been followed data were col-
centres and if the lected
type of epilepsy in
the sample match-
es the population
173

Lopez- Population of inter- low Ret-NA Variables well de- mod- Outcome is mod- NA for NA Statistical meth- mod-
Gon- est is well defined ro- scribed, pre-opera- erate briefly de- erate stud- ods well de- erate
Library
Cochrane
zalez and sample recruit- spec- tive evaluations were fined but no ies of a scribed, variables
2012 ed seems to match tive carried out in a stan- further infor- single well defined (but
this. Very clear how so dardised fashion but mation, un- group based on statisti-
sample was recruit- NA unclear how the ret- clear how out- design cal significance)
ed rospective analysis come data and results clear-
was carried out (pa- were collect- ly tabulated (un-
tient records extracted ed. Multiple clear why RRs are
Better health.
Informed decisions.
Trusted evidence.
etc.)? time points reported)
for the analy-
sis accounted
for in analysis
Mad- Descriptions of mod- Ret-NA Likely that measure- high Seizure out- mod- NA for NA Unclear what high
havan population of in- erate ro- ment of variables was come defined erate stud- kind of statistical
2007 terest and sample spec- different across the according to a ies of a model has been
recruited are not tive centres, unclear how known scale, single fitted and only
quite the same, in- so some variables have time point not group significant vari-
clusion criteria are NA been defined and stated and design ables entered in-
brief choice for analysis unclear how to the model
outcome data
was collected
McIn- Population of in- low Ret-NA Prognostic variables low Difficult out- low NA for NA Methods well mod-
tosh terest well de- ro- well defined and a come to mea- stud- described and erate
2012 scribed, study sam- spec- lot of effort has been sure in a valid ies of a results well re-
ple matches pop- tive made to accurately and reliable single ported. Howev-
ulation for charac- so collect the data for way but a lot group er, choice of vari-
teristics of inter- NA all patients; patients of effort has design ables and cate-
est, all included in came from a single been made to gories is based
analysis centre correctly clas- on the data and
sify patients statistical signifi-

and outcome cance
is measured
at one time
point in pri-
mary analysis
O'Brien Population of inter- low Ret-NA Potential for misclassi- mod- Difficult out- mod- NA for NA Statistical analy- low
2000 est is well defined ro- fication due to the na- erate come to mea- erate stud- sis described very
and sample recruit- spec- ture of the prognostic sure in a valid ies of a well with no se-
ed seems to match tive factors but efforts to and reliable single lective reporting
this. Very clear how so reduce the risk, defin- way; mea- group of results
174
NA itions well described sured ac- design

sample was recruit- but unclear why some cording to a
ed patients were not in- known scale,
Library
Cochrane
cluded in some analy- could have
ses been different
if a different
time of mea-
surement was
chosen due to
variability of
Better health.
Informed decisions.
Trusted evidence.
outcome
Paolic- Population of in- mod- Ret-NA All definitions of vari- mod- Difficult out- mod- NA for NA Unclear which high
chi terest is clear and erate ro- ables seem clear and erate come to mea- erate stud- variables have ac-
2000 sample seems to spec- little to no variation sure in a valid ies of a tually gone into
match this, howev- tive in collection/mea- and reliable single the model, what
er insufficient in- so surement methods way; unclear group "univariate con-
formation given re- NA seems to have oc- when the out- design ditions" means
garding methods of curred across patients come was and results are
recruitment but unclear if patients measured, selectively re-
came from a single probably vari- ported
centre able across
patients who
may have
come from
multiple cen-
tres
Phi Population of inter- low Ret-NA Most variables well de- mod- Difficult out- mod- NA for NA Methods section mod-
2009 est is well defined ro- fined, some uncertain- erate come to mea- erate stud- seems clear on erate
and sample recruit- spec- ty on method of measure in a valid ies of a the methods and
ed seems to match tive surement for some and reliable single factor list but un-
this. Very clear how so variables and methods way; unclear group clear from the re-
sample was recruit- NA of analysis when the out- design sults exactly what

ed come was has gone into the
measured, model
probably vari-
able across
patients
Rad- Sample recruited mod- Un-high Variables seem to be mod- Difficult out- mod- NA for NA Appropriate mod-
hakr- well described and erate clear measured in the same erate come to mea- erate stud- model, well de- erate
ishnan seems represen- if way for all patients sure in a valid ies of a scribed, howev-
1998 tative of the pop- pa- and defined clearly, and reliable single er unclear which
ulation of inter- tients some variables sub- way; could group variables had
est, however un- were jective and taken from have been dif- design been entered into
175

clear whether pa- re- patient surveys by de- ferent if a dif- the multivariate
tients were recruit- cruit- finition so could in- ferent time of model (based on
Library
Cochrane
ed prospectively or ed duce bias. Unclear measurement univariate analy-
retrospectively and prospec- how much missing da- was chosen ses?) and if all re-
how complete the tive- ta are present and how due to vari- sults have been
follow-up informa- ly this was handled in ability of out- reported
tion was or analysis come
ret-
ro-
Better health.
Informed decisions.
Trusted evidence.
spec-
tive-
ly
(or
both).
15
pa-
tients
ex-
clud-
ed
for
less
than
2
years
fol-
low-up
or
miss-
ing
da-
ta
out
of

all
con-
sec-
u-
tive
pa-
tients.
No
with-
drawals
men-
176
tioned

(but
im-
Library
Cochrane
plied),
no
fixed
study
du-
ra-
tion
Better health.
Informed decisions.
Trusted evidence.
so
com-
plet-
ing
the
study
is
NA.
Un-
clear
how
com-
plete
the
fol-
low-up
in-
for-
ma-
tion
was
and
if
this
has

im-
pact-
ed
on
analy-
sis
Rossi Population of inter- high Ret-NA Variables well defined high Difficult out- mod- NA for NA Some uncertainty high
1994 est vague and un- ro- and methods/setting come to mea- erate stud- over exactly what
clear so uncertain spec- described in some de- sure in a valid ies of a has been done in
if the sample rep- tive tail but unclear how and reliable single terms of model-
177
resents this. Little the retrospective re- way; outcome ling; variables en-

information about so view was carried out. measured group tered based on
how the sample NA A large proportion of at the same design statistical signifi-
Library
Cochrane
was recruited and missing data is likely time point cance and only P
analysis performed to have impacted on by a well-de- values reported
on a subset of this the generalisability of fined scale without interpre-
sample analyses but no details tations
regarding da-
ta collection
Better health.
Informed decisions.
Trusted evidence.
Sagher Population of inter- low Ret-NA All variables well de- low Seizure out- low Vari- mod- Statistical meth- mod-
2012 est well described, ro- scribed, multiple au- come mea- ables erate ods well de- erate
sample recruited spec- thors classified some sured ac- well scribed but un-
well described, tive variables and data cording to a de- clear which vari-
seem to match so likely to have been col- known scale fined ables went into
NA lected in the same way at specific sepa- the model and
with patients from a time points; rated how to interpret
single centre model builds by the the effect sizes
in all fol- 2 types
low-up of pa- of
tients and da- surgery,
ta collected in differ-
the same way; ences
patients from be-
a single cen- tween
tre vari-
ables
most-
ly not
tested.
Type of
surgery
includ-
ed in

GEE
model
Sarkis Population of inter- low Ret-NA All variables well de- low Seizure out- mod- NA for NA Statistical meth- high
2012 est well described, ro- scribed, data likely to come defined erate stud- ods well de-
sample recruited spec- have been collected according to a ies of a scribed, howev-
well described, tive in the same way with known scale, single er variables se-
seem to match so patients from a single not complete- group lected based on
NA centre ly clear how design statistical signif-
outcome da- icance and re-
ta was collect- sults selectively
ed, statisti- reported
178

cal model will
account for
Library
Cochrane
variable fol-
low-up length
Schramm Large sample re- mod- All low Clinical/demograph- mod- Difficult out- low Clini- mod- Unclear which mod-
2011 cruited, seems rep- erate pa- ic data seem to have erate come to mea- cal/de- erate variables have erate
resentative of the tients been collected in a re- sure in a re- mo- been included
population of inter- in liable and valid way liable and graph- in the model, ex-
Better health.
Informed decisions.
Trusted evidence.
est, however 35% the under the same pro- valid way but ic data actly what kind
of potentially eli- study tocol for all patients, seems satis- seem of analysis has
gible individuals con- however unclear what factory, mea- to have been performed
were not included tributed are prognostic vari- sured by a been and whether all
for varying reasons 1 ables and what are known scale col- results have been
year confounding factors at a single lected reported
da- time point in a re-
ta in the same liable
and way for all pa- and
were tients valid
analysed way
in under
an the
in- same
ten- proto-
tion-tocol for
treat all pa-
analy- tients,
sis. howev-
A er un-
sec- clear
ondary what
analy- are
sis prog-

was nostic
planned vari-
in ables
the and
case what
of are
drop- con-
outs found-
ing fac-
tors.
"Con-
founders"
179
don't

ap-
pear to
Library
Cochrane
be ac-
count-
ed for
in de-
sign
and
un-
Better health.
Informed decisions.
Trusted evidence.
clear
what
kind of
analy-
sis (if
any)
was
done
Spencer Large sample size, mod- Notmod- Complete data col- mod- Difficult out- mod- NA for NA Unclear whether high
2005 which is represen- erate enough
erate lected under the same erate come to mea- erate stud- the analysis was
tative of the pop- in- protocol, some scope sure in a reli- ies of a appropriate,
ulation of interest for- for variation between able and valid single which variables
but unclear how ma- patients and not all PF way; relies group were entered in-
many other eligi- tion definitions are com- on patient re- design to the model and
ble individuals de- in pletely clear porting and results are selec-
clined to take part the doesn't use a tively reported
or were excluded pub- known scale
from analyses li-
ca-
tions
on
pa-
tients

lost
to
fol-
low-up
and
not
in-
clud-
ed
in
the
analy-
180
ses

to
be
Library
Cochrane
able
to
tell
Tezer Population of in- low Ret-NA Most variables well mod- Difficult out- mod- NA for NA Unclear which high
2008 terest and sam- ro- defined, unclear how erate come to mea- erate stud- variables went
ple well described, spec- some variables de- sure in a valid ies of a into the mod-
Better health.
Informed decisions.
Trusted evidence.
sample seems to tive fined and analysed. A and reliable single el, what the out-
match population so lot of effort has been way but a lot group come variable is
of interest NA made to avoid misclas- of effort has design and results are
sification in data col- been made to selectively re-
lection and patients correctly clas- ported
came from a single sify patients
centre according
to a known
scale. Howev-
er, outcome is
measured at
variable time
points
Theodore Study sample mod- Studymod- Unclear exactly what mod- Difficult out- mod- NA for NA Unclear how the high
2012 seems to be rep- erate is erate the prognostic factors erate come to mea- erate stud- variables are be-
resentative of the prospec- of interest from the sure in a valid ies of a ing analysed in
population of in- tive PET scans are and reliable single the model and re-
terest, however with way; could group sults selectively
not clear on how 41 have been dif- design presented (no nu-
this sample was re- in- ferent if a dif- merical results at
cruited and how clud- ferent time of all)
many eligible par- ed measurement
ticipants declined pa- was chosen

to take part, if any tients, due to vari-
withdrawals oc- no ability of out-
curred etc. fixed come and not
study measured ac-
length cording to a
(du- known scale
ra-
tion
of
fol-
low-up
spec-
181

i-
fied)
Library
Cochrane
so
com-
plet-
ing
the
study
is
Better health.
Informed decisions.
Trusted evidence.
NA.
No
drop-
outs/
with-
drawals/loss-
es
to
fol-
low-up
men-
tioned
but
un-
clear
if
any
oc-
curred
Walz Sample recruited mod- No mod- Data seem to have mod- Difficult out- mod- NA for NA Statistical analy- mod-
2003 well described and erate loss-erate been collected the erate come to mea- erate stud- sis is appropri- erate
seems represen- es same way for all pa- sure in a valid ies of a ate and all re-
tative of the pop- to tients, in a way which and reliable single sults presented,

ulation of inter- fol- should minimise bias. way; could group however choice
est, however un- low-up Some variables by def- have been dif- design of variables for
clear whether pa- men- inition could be prone ferent if a dif- the multivariate
tients were recruit- tioned, to recall/misclassifica- ferent time of model based on
ed prospectively or rea- tion biases but clear measurement univariate results
retrospectively sons definitions given (ex- was chosen
for cept dichotomised due to vari-
ex- variables) ability of out-
clu- come. All pa-
sion tients from a
from single centre
analy- and outcome
182
sis measured ac-

giv- cording to a
en. known scale
Library
Cochrane
Can't
tell
if
the
study
is
of
Better health.
Informed decisions.
Trusted evidence.
a
prospec-
tive
or
ret-
ro-
spec-
tive
de-
sign
so
dif-
fi-
cult
to
judge
with-
drawals
and
drop-
outs
Wyler Population of in- mod- Out-low Variables entered in- mod- Difficult out- low Ran- low No statistical mod-
1995 terest not well de- erate come to the logistic regres- erate come to mea- domi- analysis de- erate

fined; introduction as- sion model are well sure in a valid sation scribed in the
refers more to the sess- defined and seem to and reliable has methods, only
surgical technique ment have been collected way but every bal- in the results;
than the popula- at reliably but unclear if effort seems anced unclear exactly
tion it is being ap- 1 patients were all from to be made the what has been
plied to year the sample centre and to verify the groups done, selective
is following the same seizure out- in reporting?
record- protocol comes of pa- terms
ed tients and of any
in all outcomes con-
all measured at found-
pa- the same time ing fac-
183
tients, tors

no there-
drop- fore
Library
Cochrane
outs the lo-
or gistic
miss- regres-
ing sion
da- should
ta allow
for the
Better health.
Informed decisions.
Trusted evidence.
iden-
tifica-
tion of
prog-
nostic
factors
with-
out
con-
found-
ing
Yang The study sample mod- Ret-NA Definitions and mea- mod- Difficult out- low NA for NA Unclear statisti- mod-
2011 appear to be rep- erate ro- surements of some erate come to mea- stud- cal methods erate
resentative of the spec- factors unclear sure in a re- ies of a
population of inter- tive liable and single
est but information so valid way but group
is limited NA seems satis- design
factory and
measured ac-
cording to a
known scale;
variability
in follow-up

time handled
in analysis
CFC: Complex Febrile Convusion

GEE: Generalised Estimating Equation
IED: Interictal epileptiform discharges
PF: Prognostic Factor
NA:not applicable
PET: positron emission tomography
RR: risk ratio

184

Table 6. Prognostic factors examined by 118 studies
Study MRI In- Ex- Mesial Con- His- Pre- En-
Tu- Fo-Vas- Side Uni- Post- To-
find- tracra-tent tem- cor- to- vi- cephalo-
mourcalcu- of lat- op- tal
Library
Cochrane
ings nial of po- dance ry ous ma- cor- lar re- er- er- num-
mon- re- ral of of head la- ti-mal- sec- al a- ber
i- sec- scle- pre- febrilein- cia calfor- tion or tive of
tor- tion ro- op seizures
jury dys-
ma- bi- dis- fac-
ing sis EEG pla-
tion lat- chargestors
used and sia/mal- er- record-
Better health.
Informed decisions.
Trusted evidence.
MRI for- al ed
ma- spikes
tion present
of
cor-
ti-
cal
de-
vel-
op-
ment
Aaberg 2012 X X XX 4
Adam 1996 X X X X X 5
Alfstad 2011 X X X 3
Arruda 1996 X 1
Awad 1991 X 1
Babini 2013 X X 2

Battaglia 2006 X XX X 4
Bell 2009 X 1
Berkovic 1995 X X 2
Blount 2004 X X 2
Boesebeck 2007 X XX 3
Boshuisen 2010 X X 2
185

Table 6. Prognostic factors examined by 118 studies (Continued)
Brainer-Lima 1996 X X X 3
Library
Cochrane
Cascino 1995 X 1
Chabardes 2005 X X X X 4
Chang 2009 X X X X 4
Better health.
Informed decisions.
Trusted evidence.
Chee 1993 X 1
Choi 2004a X X 2
Cossu 2005 X X X X X 5
Cossu 2008 X X X X XX X 7
Cukiert 2002 X X 2
Dagar 2011 X X X 3
Dalmagro 2005 X X X XX X X 7
Delbeke 1996 X 1
Donadio 2011 X X 2
Duchowny 1998 X X X 3
Elsharkawy 2008a X XX 3
Engman 2004 X 1

Erba 1992 X X X X 4
Erickson 2005 X X X XX X 6
Fujiwara 2012 X X X 3
Garcia 1994 X X X 3
Gelinas 2011 X 1
186

Georgakoulias 2008 X X X 3
Library
Cochrane
Gilliam 1997a X X 2
Gilliam 1997b X X X X X 5
Goldstein 1996 X X X X 4
Better health.
Informed decisions.
Trusted evidence.
Greiner 2011 X X X 3
Grivas 2006 X X X X X 5
Hajek 2009 X X 2
Hallbook 2010 X X 2
Hartzfield 2008 X X X X 4
Hartley 2002 X X 2
Hemb 2010 X X X 3
Hamiwka 2005 X X X 3
Holmes 1997 X 1
Holmes 2000 X X X 3
Sindou 2006 X X X X 4
Jack 1992 X 1

Jaramillo-Betancur 2009 X X X 3
Janszky 2003a X X X 3
Jayakar 2008 X X 2
Jehi 2012 X 1
Jennum 1993 X X X 3
187

Jeong 1999 X X X 3
Library
Cochrane
Kan 2008 X X X XX 5
Kanner 2009 X X 2
Kilpatrick 1997 X X X XX X 6
Better health.
Informed decisions.
Trusted evidence.
Kim 2009 X X X X X X 6
Kim 2010a X X X 3
Kim 2010b X X X X X X 6
Kloss 2002 X 1
Lackmayer 2013 X X 2
Lee 2006 X X 2
Lee 2008 X X X X 4
Lee 2011 X X X 3
Li 1997 X X XX 4
Li 1999 X X XX 4
Liang 2012 X 1
Liava 2012 X X X X X 5

Lorenzo 1995 X 1
Madhavan 2007 X
Mathern 1999 X X XX 4
Mihara 2004 X X 2
Miserocchi 2013 X X X X X X 6
188

Morino 2009 X 1
Library
Cochrane
Morris 1998 X X X 3
O'Brien 1996 X 1
O'Brien 2000 X X X X X XX X 8
Better health.
Informed decisions.
Trusted evidence.
Paglioli 2006 X 1
Paolicchi 2000 X X X X 4
Perego 2009 X X X X X X 6
Perry 2010 X X X X XX XX 8
Prevedello 2000 X 1
Raabe 2012 X X X 3
Radhakrishnan 1998 X X X X 4
Remi 2011 X 1
Rossi 1994 X X X X X 5
Sagher 2012 X 1
Sakamoto 2009 X X X X X 5
Schramm 2011 X X X 3

Seymour 2012 X X 2
Sinclair 2003 X X X X X 5
Sola 2005 X X X 3
Spencer 2005 X X X X X 5
Sperling 1992 X 1
189

Stavrou 2008 X 1
Library
Cochrane
Swartz 1992 X X X 3
Tanriverdi 2010 X X 2
Tatum 2008 X X X X 4
Better health.
Informed decisions.
Trusted evidence.
Terra-Bustamante 2005a X X X X X X XX 8
Tezer 2008 X X X 3
Theodore 2012 X X X X X 5
Tripathi 2008 X 1
Urbach 2007 X XX 3
Walz 2003 X X X X X 5
Weinand 1992 X 1
Widdess-Walsh 2007 X X X 3
Wray 2012 X X X XX X 6
Wyler 1995 X 1
Wyllie 1998 X X X X 4
Yang 2011 X X X X XX XX 8

Yeon 2009 X 1
Yu 2012a X 1
Yu 2012b X 1
Zentner 1995 X X X 3
Zentner 1996 X X 2
190

Informed decisions.


6
21
37
4519
4 41
7
15
23
46
41
21
42
Total number of studies contributing to each factor

Informed decisions.


APPENDICES
Appendix 1. Cochrane Epilepsy Group Specialised Register search strategy

#1 MeSH DESCRIPTOR Neurosurgical Procedures Explode All WITH AE CL CT EC ED ES HI IS LJ MT MO NU PX RH ST SN TD UT VE
#2 (surger* OR surgic*):TI OR (surger* OR surgic*):AB OR (surger* OR surgic*):KW OR (surger* OR surgic*):KY
#3 4/01/2013 TO 4/07/2013:CRSCREATED
#4 (#1 OR #2) AND #3
Appendix 2. CENTRAL search strategy

#1 (epilep* NEAR/4 surg*):ti,ab
#2 (extratemporal NEAR/4 surg*):ti,ab
#3 (temporal NEAR/4 surg*):ti,ab
#4 (TLE NEAR/4 surg*):ti,ab
#5 (#1 OR #2 OR #3 OR #4)
#6 (epilep* or seizure* or convuls*):ti,ab
#7 MeSH descriptor Epilepsy explode all trees
#8 MeSH descriptor Seizures explode all trees
#9 (#6 OR #7 OR #8)
#10 MeSH descriptor Neurosurgical Procedures explode all trees
#11 (#9 AND #10)
#12 (#5 OR #11)
Appendix 3. MEDLINE search strategy

This strategy is based on the Cochrane Highly Sensitive Search Strategy for identifying randomised trials (Lefebvre 2011).
1. (epilep$ adj4 surg$).ti,ab.
2. ((extratemporal or temporal or TLE) adj4 surg$).ti,ab.
3. exp Epilepsy/
4. exp Seizures/
5. (epilep$ or seizure$ or convuls$).ti,ab.
6. 3 or 4 or 5
7. exp Neurosurgical Procedures/
8. 6 and 7
9. 1 or 2 or 8
10. (english or french or german or italian or spanish).lg.
11. 9 and 10
12. exp animals/ not humans.sh.
13. 11 not 12

Informed decisions.

14. randomized controlled trial.pt.
15. controlled clinical trial.pt.
16. randomized.ab.
17. clinical trials as topic.sh.
18. randomly.ab.
19. trial.ti.
20. exp Case-Control Studies/
21. Cohort Studies/
22. (case control*).tw.
23. retrospective.tw.
24. (observational adj (study or studies)).tw.
25. 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21 or 22 or 23 or 24
26. 13 and 25
Appendix 4. EPHPP Quality Assessment Tool

QUALITY ASSESSMENT TOOL FOR QUANTITATIVE STUDIES
COMPONENT RATINGS
A) SELECTION BIAS
(Q1) Are the individuals selected to participate in the study likely to be representative of the target population?
1 Very likely 2 Somewhat likely 3 Not likely 4 Can't tell
(Q2) What percentage of selected individuals agreed to participate?

1 80% to 100% agreement 2 60% to 79% agreement 3 less than 60% agreement 4 Not applicable 5 Can't tell
RATE THIS SECTION STRONG MODERATE WEAK

See dictionary 1 2 3
B) STUDY DESIGN
Indicate the study design
1 Randomised controlled trial

2 Controlled clinical trial
3 Cohort analytic (two group pre + post)
4 Case-control
5 Cohort (one group pre + post (before and after))
6 Interrupted time series
7 Other specify ____________________________
8 Can't tell
Was the study described as randomised? If NO, go to Component C.

No Yes
If Yes, was the method of randomisation described? (See dictionary)
No Yes
If Yes, was the method appropriate? (See dictionary)
No Yes
C) CONFOUNDERS
(Q1) Were there important differences between groups prior to the intervention?
Informed decisions.

1 Yes
2 No
3 Can't tell
The following are examples of confounders:

1 Race
2 Sex
3 Marital status/family
4 Age
5 SES (income or class)
6 Education
7 Health status
8 Pre-intervention score on outcome measure
(Q2) If yes, indicate the percentage of relevant confounders that were controlled (either in the design (e.g. stratification, matching) or
analysis)?
1 80% to 100% (most)
2 60% to 79% (some)
3 Less than 60% (few or none)
4 Can't tell

D) BLINDING
(Q1) Was (were) the outcome assessor(s) aware of the intervention or exposure status of participants?
1 Yes
2 No
3 Can't tell
(Q2) Were the study participants aware of the research question?

1 Yes
2 No
3 Can't tell

E) DATA COLLECTION METHODS
(Q1) Were data collection tools shown to be valid?

1 Yes
2 No
3 Can't tell
(Q2) Were data collection tools shown to be reliable?

1 Yes
2 No
3 Can't tell

F) WITHDRAWALS AND DROP-OUTS
(Q1) Were withdrawals and drop-outs reported in terms of numbers and/or reasons per group?
1 Yes
2 No
3 Can't tell
4 Not applicable (i.e. one time surveys or interviews)
(Q2) Indicate the percentage of participants completing the study. (If the percentage differs by groups, record the lowest).
1 80% to 100%
2 60% to 79%
Informed decisions.

3 Less than 60%

4 Can't tell
5 Not applicable (i.e. retrospective case-control) RATE THIS SECTION

G) INTERVENTION INTEGRITY
(Q1) What percentage of participants received the allocated intervention or exposure of interest?
1 80% to 100%
2 60% to 79%
3 Less than 60%
4 Can't tell
(Q2) Was the consistency of the intervention measured?

1 Yes
2 No
3 Can't tell
(Q3) Is it likely that subjects received an unintended intervention (contamination or co-intervention) that may influence the results?
4 Yes
5 No
6 Can't tell
H) ANALYSES
(Q1) Indicate the unit of allocation (circle one)
Community organisation/institution practice/office individual
(Q2) Indicate the unit of analysis (circle one)
Community organisation/institution practice/office individual
(Q3) Are the statistical methods appropriate for the study design?
1 Yes
2 No
3 Can't tell
(Q4) Is the analysis performed by intervention allocation status (i.e. intention-to-treat) rather than the actual intervention received?
1 Yes
2 No
3 Can't tell
GLOBAL RATING
COMPONENT RATINGS
Please transcribe the information from the A-F boxes on pages 1-4 into this section.
A SELECTION BIAS STRONG MODERATE WEAK
123
B STUDY DESIGN STRONG MODERATE WEAK
123
C CONFOUNDERS STRONG MODERATE WEAK
123
D BLINDING STRONG MODERATE WEAK
123
E DATA COLLECTION METHOD STRONG MODERATE WEAK
123
F WITHDRAWALS AND DROP-OUTS STRONG MODERATE WEAK
123
GLOBAL RATING FOR THIS PAPER (circle one):

Informed decisions.

1 STRONG (no WEAK ratings)

2 MODERATE (one WEAK rating)
3 WEAK (two or more WEAK ratings)
With both reviewers discussing the ratings:

Is there a discrepancy between the two reviewers with respect to the component (A-F) ratings?
No Yes
If yes, indicate the reason for the discrepancy

1 Oversight
2 Differences in interpretation of criteria
3 Differences in interpretation of study
Final decision of both reviewers (circle one): 1 STRONG

2 MODERATE
3 WEAK
CONTRIBUTIONS OF AUTHORS
SW, RN, SN and JW wrote the study protocol and text of this review.
SW, RN and SN assessed potential studies for inclusion. SW, RN, JC, AS and SG reviewed the papers and abstracted data. RR performed
data extraction for four papers published in Spanish.
SN created a Microsoft database for data extraction, double-checked extracted data relating to prognostic factors and performed data
analysis.
SW, RN, SN and JW performed independent quality and 'Risk of bias' assessment of all included studies.
DECLARATIONS OF INTEREST
SW has no interests to declare
SN has no interests to declare
JC has no interests to declare
SG has no interests to declare
AS has no interests to declare
JW has no interests to declare
RR has no interests to declare
RN has no interests to declare
SOURCES OF SUPPORT
Internal sources
• No sources of support supplied
External sources
• National Institute for Health Research, UK.
This review was supported by the National Institute for Health Research, via Cochrane Programme Grant funding to the Epilepsy Group.
The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the Systematic Reviews
Programme, NIHR, NHS or the Department of Health.
DIFFERENCES BETWEEN PROTOCOL AND REVIEW
We specified in the protocol, before the identification of eligible studies, that we would summarise outcomes by pooled odds ratios (OR)
and 95% confidence intervals. At the time of review, once all eligible studies had been identified, we felt that summarising outcomes by
pooled risk ratios (RR) and 95% CIs would be more appropriate for the design of the included studies due to the ease of interpretation of

Informed decisions.

RRs compared to ORs. This approach is often more easily interpretable by clinicians and allows them more readily to translate study results
into useful information for future surgical candidates and their families who are facing difficult decisions.
We originally defined one of our prognostic factors of interest as "presence of neuromigrational defect" in concordance with the prognostic
factors defined in Tonini 2004; we have adapted this definition to "presence of focal cortical dysplasia or malformation of cortical
development" to better reflect the pathologies included in this prognostic factor and to avoid confusion or misinterpretation.
We specified in the protocol that adjusted (multivariable) and unadjusted (univariate) analyses of prognostic factors would be compared,
however insufficient data were available from published papers to allow this comparison. We are aware that presenting a univariate
analysis alone in this review does not give a full account of the effect of prognostic factors, which are unlikely to act independently upon
outcome. Following advice from the Cochrane Prognostic Methods Group, we have added an extra section to the review, which narratively
summarises the results of 29 studies that fit multivariable prognostic regression models including the factors of interest in this review. We
also performed an additional post hoc risk of bias assessment using the Quality of Prognostic Studies (QUIPS) tool on the 29 studies.
INDEX TERMS
Medical Subject Headings (MeSH)

Analysis of Variance; Anticonvulsants [therapeutic use]; Epilepsies, Partial [drug therapy] [*surgery]; Randomized Controlled Trials as
Topic; Treatment Outcome
MeSH check words

Humans


West Et Al-2015-Cochrane Database of Systematic Reviews

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

West Et Al-2015-Cochrane Database of Systematic Reviews

Uploaded by

Copyright:

Available Formats

West S, Nolan SJ, Cotton J, Gandhi S, Weston J, Sudan A, Ramirez R, Newton R

West S, Nolan SJ, Cotton J, Gandhi S, Weston J, Sudan A, Ramirez R, Newton R.

Surgery for epilepsy (Review) i

Surgery for epilepsy (Review) ii

Surgery for epilepsy

Editorial group: Cochrane Epilepsy Group.

Data collection and analysis

Surgery for epilepsy (Review) 1

Surgery for epilepsy

Surgery for epilepsy (Review) 2

Surgery for epilepsy (Review) 3

Settings: outpatients (following surgery in hospital)

Comparison: medical treatment

Medical treat- Surgery

Cochrane Database of Systematic Reviews

Comparison of surgical interventions for epilepsy

Intervention: experimental surgical intervention (see comments)

Comparison: control surgical intervention (see comments)

Outcomes Illustrative comparative risks* (95% CI) Relative No of Qual- Comments

surgical in- surgical intervention

without corpus callosotomy (ATL)

Cochrane Database of Systematic Reviews

GRADE Working Group grades of evidence

Cochrane Database of Systematic Reviews

BACKGROUND complex in short-term memory, consciousness and emotions. As

Surgery for epilepsy (Review) 7

Class 1: Free of disabling 1A: Completely seizure-free since surgery

Class 3: Worthwhile improve- 3A: Worthwhile seizure reduction

Class 4: No worthwhile im- 4A: No significant seizure reduction

Surgery for epilepsy (Review) 8

• Postoperative discharges: presence or absence of EEG

Surgery for epilepsy (Review) 9

Electronic searches 6. prognostic indicators: different indicators are described as

Surgery for epilepsy (Review) 10

Data synthesis Instead, as described in Subgroup analysis and investigation of

Surgery for epilepsy (Review) 11

RESULTS From searches conducted by the Cochrane Epilepsy Group between

Surgery for epilepsy (Review) 12

Figure 1. Study flow diagram.

Surgery for epilepsy (Review) 13

We excluded a total of 93 studies from the review. We excluded A. Selection bias

Surgery for epilepsy (Review) 15

Surgery for epilepsy (Review) 16

Surgery for epilepsy (Review) 17

H. Analyses Despite the described applicability of this tool to all quantitative

Surgery for epilepsy (Review) 18

Surgery for epilepsy (Review) 19

In Wyler 1995, at one year, 25 out of 36 (69%) of the participants

Surgery for epilepsy (Review) 20

Surgery for epilepsy (Review) 21

Surgery for epilepsy (Review) 22

Surgery for epilepsy (Review) 23

Surgery for epilepsy (Review) 24

Surgery for epilepsy (Review) 25

Surgery for epilepsy (Review) 26

Surgery for epilepsy (Review) 28

Surgery for epilepsy (Review) 29

Surgery for epilepsy (Review) 30

Surgery for epilepsy (Review) 32

Surgery for epilepsy (Review) 35

Surgery for epilepsy (Review) 37

Surgery for epilepsy (Review) 40

Surgery for epilepsy (Review) 41

Surgery for epilepsy (Review) 42