Temporal lobe epilepsy surgery and the quest for optimal extent of resection: A review


Address correspondence to Johannes Schramm, M.D., Department of Neurosurgery, University of Bonn, Sigmund-Freud-Str. 25, D-53105 Bonn, Germany. E-mail: Johannes.Schramm@ukb.uni-bonn.de


The efficacy of surgery to treat drug-resistant temporal lobe epilepsy (TLE) has been demonstrated in a prospective randomized trial. It remains controversial which resection method gives best results for seizure freedom and neuropsychological function. This review of 53 studies addressing extent of resection in surgery for TLE identified seven prospective studies of which four were randomized. There is considerable variability between the intended resection and the volumetrically assessed end result. Even leaving hippocampus or amygdalum behind can result in seizure freedom rates around 50%. Most authors found seizure outcome in selective amygdalohippocampectomy (SAH) to be similar to that of lobectomy and there is considerable evidence for better neuropsychological outcome in SAH. Studies varied in the relationship between extent of mesial resection and seizure freedom, most authors finding no positive correlation to larger mesial resection. Electrophysiological tailoring saw no benefit from larger resection in 6 of 10 studies. It must be concluded that class I evidence concerning seizure outcome related to type and extent of resection of mesial temporal lobe structures is rare. Many studies are only retrospective and do not use MRI volumetry. SAH appears to have similar seizure outcome and a better cognitive outcome than TLR. It remains unclear whether a larger mesial resection extent leads to better seizure outcome.

The clinical entity of temporal lobe epilepsy (TLE) may be considered fairly well defined. The large majority of TLE originated from the mesiotemporal part of the lobe and in 1993 Wieser et al. defined the clinical picture of mesial TLE (MTLE) (Wieser et al., 1993). For many years temporal lobe resections (TLRs) have been the most frequently used procedure for TLE. That a distinction can be made between MTLE and lateral neocortical temporal lobe epilepsy (NTLE) is a more recent development and the efficacy of this hypothesis has been demonstrated in series with resections confined to the lateral part of the temporal lobe (Schramm et al., 2001; Janszky et al., 2006). The role of the neurosurgeon in the surgical treatment of TLE has evolved from doing so-called standard anterior lobectomies (TLR) to more restricted variants of anterior lobectomy such as the so-called Spencer type of resection, which consists of a small (ca. 3 cm) anterior partial lobectomy combined with a more extensive mesial resection (Spencer et al., 1984). After Niemeyer's (1958) early transcortical amygdalohippocampectomy (Niemeyer 1958), Yasargil Wieser have proposed a strictly mesial resection type—transsylvian selective amygdalohippocampectomy (SAH) as treatment for MTLE (Niemeyer, 1958; Yasargil et al., 1985; Yasargil et al., 1993).

Since Wyler's randomized study on mesial extent of resection, his statement that there is “a growing opinion, that optimal seizure outcome from TLR can be expected if the majority or all of the hippocampus (HC) is removed” is frequently used in discussions of TLE surgery (Bonilha et al., 2007). Since Wiebe et al., (2001) demonstrated that TLR is more effective than medical management in the treatment of drug-resistant TLE, the question of the extent of resection has grown in importance.

Neuropsychological deficits in patients with chronic TLE or MTLE are frequently aggravated following surgery (Milner, 1958; Milner, 1972; Hermann et al., 1992; Helmstaedter et al., 1996; Jones-Gotman et al., 1997; Helmstaedter & Kurthen, 2001). Although in a proportion of patients postoperative cognitive improvement has been observed, postoperative decline in cognitive function following TLE surgery has always been a concern for epileptologists and neurosurgeons. This has been the reason to try to minimize adverse neuropsychological outcomes by using more conservative neurosurgical approaches and less extensive resections (Jack et al., 1988). Obviously, demonstration of the superiority of such an approach would require to demonstrate that despite smaller extents of resection seizure outcome remains as good as with more extensive resection types. The interpretation of studies comparing resection extent or resection types is hampered by several facts. Most were performed in a retrospective fashion, and few were controlled postoperatively by MRI for precise extent of resection (Jones-Gotman et al., 1997). It is also remarkable how variable the extent of resection can be, both between patients from one center and between different centers.

The question of the effect of extent of resection on seizure and neuropsychological outcome should be considered on two levels. The first level is the decision of the surgeon to remove certain structures to a defined degree using one of several standardized approaches. The second and far less addressed question is that of how much of the intended resection volume was really removed (Jack et al., 1988; Awad et al., 1989; Siegel et al., 1990; Jones-Gotman et al., 1997; Vajkoczy et al., 1998).

This paper reviews studies on surgical outcome as a function of the type and extent of resection with special consideration of the study design (prospective/retrospective) and the question of whether MRI control of extent of resection had been performed. It was not the aim of this study to perform a metaanalysis and to consider all possible factors influencing outcome in epilepsy surgery as this has been done recently (Tonini et al., 2004). This paper focuses on aspects, which are relevant for the neurosurgeon, e.g., the variability of surgical approaches. The neurosurgeon is responsible for what is done in the operating room, although the type and extent of resection is usually agreed on in the epilepsy case conference between the epileptologist and the neurosurgeon. The epilepsy surgery community should therefore recognize the relationship between extent of resection and outcome and the degree of variability in extent of resection in so-called standard resection types.


Publications were identified not by a systematic search of bibliography databases using specific keywords but through the author's initial identification of relevant articles and then a search of related articles in the PubMed database. Reviewing the status of knowledge on the optimal extent of resection, several study types can be identified. Seven earlier studies were devoted to the question whether outcome is satisfactory if only lateral temporal lobe parts were removed with amygdalum (AM) left behind or even with amygdalum and HC left behind (Feindel & Rasmussen, 1991; Leonard, 1991; Loring et al. 1991; Olivier, 1991; Goldring et al., 1992; Keogan et al., 1992; Jooma et al., 1995). Another type of study was devoted to the comparison of two resection types, frequently SAH versus lobectomy (Tables 1 and 2). Three studies were prospective (Awad et al., 1989; McKhann et al., 2000; Helmstaedter et al., 2008), four were randomized (Wyler et al., 1995; Hermann et al., 1999a; Hadar et al., 2001; Lutz et al., 2004), most of them retrospective. Another study type was devoted to a retrospective analysis of possible predictors for good outcome in relation to extent of resection. In a subgroup of the lobectomy studies (n = 10), the extent of resection was determined by intraoperative ECoG mapping, or extent of atrophy on preoperative MRI. Typically all of these studies looked at seizure outcome, over 20 also looked at the effects on postoperative neuropsychological outcome, 15 of these comparing SAH with TLR.

Table 1.  Studies comparing TLR with SAH for seizure outcome
Arruda et al. (1996)Similar
Clusmann et al. (2002)Similar in adults
Hadar et al. (2001)Similar
Paglioli et al. (2006)Similar
Pauli et al. (1999)Similar
Tanriverdi & Olivier (2007)Similar
Clusmann et al. (2004)Better in TLR in children + adolescents
Bate et al. (2007)Better in TLR
Renowden et al. (1995)Same outcome for two types of SAH
Lutz et al. (2004)Same outcome for two types of SAH
Table 2.  Studies comparing TLR with SAH for neuropsychological outcome
Clusmann (2004) SAH in adults: higher rate of improvement + lower rate of deterioration in overall neuropsychological score
Goldstein & Polkey (1993) SAH: short-term beneficial effect on memory
Hadar et al. (2001) SAH better in recall, no difference in Boston naming test or WMS-R test
Helmstaedter et al. (1996) Immediate recall better in SAH
Helmstaedter et al. (2002) SAH has advantage over TLR in long-term follow-up (2–10 years)
Helmstaedter et al. (2008) SAH better for R-sided resection, TLR better for L-sided resection for material-specific memory
Morino et al. (2006) SAH better memory function
Paglioli et al. (2006) SAH better for verbal memory score (30% deterioration in both groups)
Pauli et al. (1999) SAH better for verbal memory (no change)
Renowden et al. (1995) SAH better in verbal IQ and nonverbal memory
Tanriverdi & Olivier (2007) SAH less decline for verbal memory
Hader et al. (2005) No difference
Wolf et al. (1993) No difference
Goldstein & Polkey (1992) No difference when using global memory test
Jones-Gotman (1997) Similar deficits in learning + retention tasks in seizure free patients


Resection or sparing of mesial structures

Several studies were devoted to the question of which particular subcompartment of the mesial structures affects outcome, i.e., what effect did it have if the AM was resected or not, if the HC was resected anteriorly or more posteriorly during TLR (Feindel & Rasmussen, 1991; Leonard, 1991; Loring et al., 1991; Olivier, 1991; Goldring et al., 1992; Keogan et al., 1992). In the early 1990s several authors described their results in TLE surgery leaving structures behind, which are today considered necessary to achieve seizure freedom. Feindel & Rasmussen (1991) did not remove the HC in 100 cases and achieved 53% seizure freedom and 63% satisfactory outcome. Goldring et al. (1992) did not remove the AM in 70 cases but included 2.5 cm of the HC and achieved a 60% (class I plus class II) satisfactory seizure outcome similar to cases where the AM was also removed. Keogan et al. (1992) reported on 50 cases where he left AM and HC behind and achieved 46% seizure freedom. Half of his cases had postoperative MRI and he concluded that a greater volume of lateral tissue removal predicted a good result. Kanner et al. (1995)—using electrophysiological tailoring—found that sparing of HC or HC plus AM is not necessarily related to poor outcome. Loring et al. (1991), comparing cases with anterior hippocampal resection with cases with hippocampal sparing, described that a more extensive and more posterior mesial temporal resection is not necessarily associated with a greater verbal memory deficit. Jooma et al. (1995) in 70 cases where the extent of resection was determined by interictal seizure activity recorded with depths electrodes achieved 60% seizure freedom if only the AM was resected, 60% seizure freedom if 1–2 cm of HC was also resected, and 78% with more than 2 cm of the HC resected. Olivier (1991) reported on 241 cases comparing right and left TLR with and without hippocampectomy considering the two sides separately. He found that the failure rate was higher when the HC has been left alone and achieved seizure freedom in 48.5% of left TLR including hippocampectomy versus 34% without hippocampectomy, for the right side 67% with hippocampectomy versus 60% without hippocampectomy. In the same year also from Montreal Leonard (1991) reported on cases operated under Penfield's and Feindel's rules, that there was no difference in outcome for the various degrees of HC resection in TLR. In summary it appears that leaving behind one of the two main mesial components (AM or HC) totally or to a large part leads to somewhat poorer outcomes, but not to dismally poor results. Nearly all of these series did not use preoperative or postoperative MRI and all studies were retrospective.

Lobectomy versus SAH: Seizure outcomes

A number of studies were devoted to the comparison of seizure results in temporal lobectomy (TLR) with SAH (Arruda et al., 1996; Jones-Gotman et al., 1997; Pauli et al., 1999; Hadar et al., 2001; Clusmann et al., 2002; Clusmann et al., 2004; Paglioli et al., 2006; Bate et al., 2007; Tanriverdi & Olivier 2007) (Table 1). In addition two studies compared two forms of SAH with each other (Renowden et al., 1995; Lutz et al., 2004). Seizure outcome was similar in adults in a large retrospective analysis of 126 SAH cases versus 83 TLR cases (Clusmann et al., 2002) but worse for SAH in children (Clusmann et al., 2004). Seizure outcome was similar in the 28 patients reported by Hadar et al., in the 69 patients comparing three resection types by Pauli et al. (1999), in the 161 patients by Paglioli et al. (2006), in the 74 patients by Arruda et al. (1996), and in a series of 72 patients from Montreal comparing corticoamygdalohippocampectomy with SAH (Tanriverdi & Olivier, 2007). The two studies comparing transsylvian with transcortical SAH (Renowden et al., 1995; Lutz et al., 2004) also had similar seizure outcomes for the two approaches. Thus, most of these authors came to the conclusion that seizure outcome was similar in both resection types (Table 1), only two authors pointed out that seizure outcome was different depending on resection type with one author finding better outcomes after TLR (Bate et al., 2007) and the other finding better outcomes after TLR in children only (Clusmann et al., 2004). In other words, six of eight studies did not demonstrate a worse seizure outcome for the smaller resection type “SAH.”

Lobectomy versus SAH: Neuropsychological outcomes

Comparing neuropsychological sequelae or outcomes for SAH and TLR (Table 2) several authors have found a better neuropsychological outcome for SAH. Goldstein & Polkey (1993) found a short-term beneficial effect on memory for SAH (Goldstein & Polkey, 1993) but could not demonstrate the same effect when using a more global test of memory (Goldstein & Polkey, 1992). Siegel et al. (1990) in a small sample of their patients described memory benefits after the selective procedure. Clusmann et al. (2002) showed a low rate of improvement and high rate of deterioration on overall neuropsychological score for TLR compared to SAH and lesionectomy. Paglioli et al. (2006) found that SAH is possibly better for verbal memory score, although 30% of all patients in both groups deteriorated independent of surgical approach. (Morino et al. 2006) found memory function better preserved after SAH and Pauli et al. (1999) found that tailored resections did better than classical lobectomies and that the type of resection influenced verbal memory performance postoperatively with no change in verbal memory in SAH cases. Renowden et al. (1995) found that left SAH patients did better in verbal IQ and nonverbal memory than those with left TLR. Clusmann et al. (2002) demonstrated with left-sided surgery that memory decline was more frequent in lobectomy (43%) than in SAH (30%). Hadar et al. (2001) found an advantage for SAH in recall but no difference in Boston naming or WMS-R tests. Helmstaedter et al. (1996) also found an advantage for immediate recall for SAH cases. In a prospective study comparing SAH with anterior one-third lobectomy plus mesial resection (Helmstaedter et al., 2008), SAH appeared to be favorable for material-specific memory in right-sided resections and combined pole plus mesial resection more favorable in left-sided surgeries. Comparing SAH with corticoamygdalohippocampectomy Tanriverdi & Olivier (2007) described a lesser decline in verbal memory after left-sided SAH. Loring et al. (1991) found SAH better for material-specific memory with right-sided SAH but not for left-sided SAH. Wolf et al. (1993) found no cognitive differences between groups defined by extent of mesial and lateral resection (Wolf et al., 1993). Hader et al. (2005) also found no differences in 36 cases (Hader et al., 2005). Jones-Gotman et al. (1997) compared seizure free cases after three resection types (SAH, TLR, and lateral neocorticectomy) and found very similar deficits in learning and retention tasks in all three groups. Helmstaedter et al. (1996) also compared these three resection types and found an advantage of SAH over TLR and the best outcome for pure cortical lesionectomy. Using the same outcome measures an advantage of selective AH to TLR was also seen in a longitudinal follow-up with intervals ranging from 2 to 10 years after surgery (Helmstaedter et al., 2002). In summary there is considerable evidence for somewhat better neuropsychological results with SAH, although this was not consistently found.

Extent of resection in TLR and seizure outcome

Another type of study compares differing extents of resection, sometimes only focusing on mesial differences, sometimes looking at both mesial and lateral differences in extent of resection (Table 3), several studies used postoperative MRI to assess extent of resection (Table 3). Three studies compared extent of mesial resection in SAH (Vajkoczy et al., 1998; Abosch et al., 2002; van Rijckevorsel et al., 2005) mostly using postoperative MRI. Major weakness in these studies are that few are prospective and there is a lack of postoperative MRI data: two studies were prospective and randomized, one of which did not control postoperative extent of resection by MRI (Wyler et al., 1995) the other one is not fully published (Hadar et al., 2001).

Table 3.  Studies with extent of resection as seizure outcome factor
Bonilha et al. (2007)MRI HC and entorhinal cortex resection correlate with better outcome, but not overall size of resection
Nayel et al. (1991)MRI Larger extent has positive influence
Siegel et al. (1990)MRI PHG extent correlates with better outcome, general tendency for better outcome with longer resection
Wyler et al. (1995) Complete HC resection—much better
Bonilha et al. (2004)MRI Extent of HC resection associated with outcome
Arruda et al. (1996) Similar for SAH (smaller resection) + TLR (larger)
Jack et al. (1988)MRI Total amount not necessarily correlated
 Jones-Gotman et al. (1997)MRI Significally different removal extents of HC and AM compatible with seizure freeness
Kanner et al. (1995) Limited HC resection not related to poor outcome
McKhann et al. (2000) Extent need not be maximal if ECoG tailoring on HC is used
Renowden et al. (1995)MRI HC + AM extent not directly related
Son et al. (1994) Similar outcome (>2 cm vs. <2 cm HC resection)
Van Rijckevorsel et al. (2005)MRI HC extent not correlated (in SAH)

Five studies found that a larger extent of resection positively influenced seizure outcome. Nayel et al. (1991) in a study where each patient had postoperative MRI came to the conclusion that mesial extent of resection was the most important factor and Wyler et al. (1995) found a much better seizure outcome for extensive mesial resection in his prospective randomized study, with the same neuropsychological outcome. However, in that study there was no postoperative MRI control of true extent of resection. Siegel et al. (1990) found a general tendency in a MRI control study for a better outcome to be obtained from a larger resection and described that the extent of parahippocampal gyrus (PHG) resection correlated with good seizure outcome. Bonilha et al. (2004) compared good outcome with unsatisfactory outcome and found a significant association between MRI-measured extent of HC removal and outcome. Bonilha et al. (2007) in a voxel-based MRI study found that extent of resection for HC and entorhinal cortex showed a significant linear correlation with better seizure outcome, but that overall size of resection did not correlate. Contrary to Siegel's results (1990) entorhinal cortex rather than PHG resection removal was important. The majority of authors conclude that the extent of mesial or lateral resection is not important for outcome. Wolf et al. (1993) found no statistically significant difference in seizure outcome for smaller versus larger mesial resections having formed groups with a mesial resection of less than 2 cm or more than 2 cm and lateral resection smaller than 4 cm or larger than 4 cm. For the two larger resection groups seizure outcome was also similar. The neuropsychological outcome was described as showing no differences in cognition between the groups defined by extent of resection, both laterally and mesially. Loring et al. (1991) compared lobectomies with anterior resection versus lobectomies with anterior hippocampal sparing. He concluded that more extensive and more posteriorly resection was not necessarily associated with more verbal memory deficit but did not comment on seizure outcome. Hermann et al. (1999b) defined four different types of TLR (with and without superior temporal gyrus (STG) resection, with and without tailoring by mapping) and did not discover differences in visual confrontation naming decline between the groups which was found in all four groups regardless of technique. Renowden et al. (1995) and van Rijckevorsel et al. (2005) found in a MRI control study of SAHs that the extent of hippocampal and AM resection did not seem to be directly related to seizure cure. Jennum et al. (1993) grouped the extent of neocortical TLR into larger or smaller than 5 cm guided by electrocorticography (ECoG) and found no difference in seizure outcome for the two groups. Jack et al. (1988) in a MRI control study described that the total amount of tissue resection did not necessarily correlate with clinical outcome in individual cases. Arruda et al. (1996) described that the degree of atrophy was more important for seizure outcome than the type of resection. Kanner et al. (1995) found that the sparing or a more limited resection of HC is not necessarily related to a poor outcome. Jones-Gotman et al. (1997) looked at seizure free cases and compared three different resection types from three institutions (lesionectomy, SAH, and TLR) and came to the conclusion that considerable differences in removal extent of HC and AM are compatible with a seizure free outcome. Son et al. (1994) tailored the extent of mesial resection in two groups (smaller and larger than 2 cm hippocampectomy) in TLR and found the same seizure outcome in both groups. In summary, five series found better seizure outcomes with more extensive resection, but seven studies did not.

Extent of resection in TLR and neuropsychological outcome

Wolf et al. (1993) described that the risk of cognitive impairment depends more on the age at seizure onset, than on resection extent. Resection of nonatrophic HC was associated with a poorer verbal memory outcome in Trenerry's study, (Trenerry et al., 1993) i.e., degree of atrophy appeared more relevant than extent of resection. Loring et al. (1991) found that a larger and more posterior mesial resection was not associated with more verbal memory deficit. Leonard (1991) compared several groups with varying extent of the removal from the early Montreal series and did not find neuropsychological differences between them. Wyler et al. (1995) saw no increased neuropsychological morbidity with more extensive HC resection in their randomized TLR series. Hermann et al. (1999a) found in a prospective study that resection of nonsclerotic HC, used as a synonym for late age onset TLE was associated with nominal speech decline, whereas preservation or resection of the STG did not influence these factors (Hermann et al., 1999a). The same group retrospectively compared neuropsychological impairment in four different types of TLR (with and without STG sparing and with or without intraoperative mapping) and found no difference in decline of visual confrontation naming (Hermann et al., 1999b). Jones-Gotman et al., (1997) in their comparison of three resection types, comparing only seizure free patients, described in a MRI control study that similar deficits of learning and retention tasks for abstract words and designs was seen in all three resection groups and also similar for small and larger extent of the three important mesiobasal compartments. Katz et al. (1989) examined 20 patients with the 20-compartment model and described that postoperative decrease in percentage of retention of verbal material correlated with extent of mesial resection in left TLR (Katz et al., 1989). Alpherts et al. (2007) examined cognitive outcomes in standard and tailored TLR, controlling for extent of three lateral gyri: A larger STG resection in standard TLR correlated with a decrease in verbal memory, tailored TLR showed a decline in short-term memory and attention (Alpherts et al., 2007 in press, corrected proof). So one study clearly declares that the extent of mesial resection in TLR is important for neuropsychological outcome, eight did not see larger deficits with larger resection, and one found larger declines in tailored (i.e., smaller) resections. When comparing these studies one must keep in mind that somewhat different outcomes may well result from application of different neuropsychological tests. Test procedure, test materials, as well as application of parallel tests, and repeated use of the same test can well be discussed as a complicating factor which makes it difficult to come to a final conclusion.

Tailoring of extent of resection with electrophysiology: Seizure and neuropsychological outcome

Another group of studies looked at different outcomes following intraoperative tailoring based on electrophysiological recordings, representing a nonmorphological approach (Ojemann & Dodrill, 1985; Jennum et al., 1993; Son et al., 1994; Cascino et al., 1995; Jooma et al., 1995; Kanner et al., 1995; McKhann et al., 2000; Paolicchi et al., 2000; Leijten et al., 2005).

Jooma et al. (1995) determined extent of resection of HC by activity with depth electrodes and found better neuropsychological results if no HC was resected and concluded that a pure AM resection may be sufficient. Son et al. (1994) tailored the size of the hippocampectomy by ECoG and formed two groups with more than and less than 2 cm hippocampectomy and found no difference in postoperative neuropsychology. Ojemann & Dodrill (1985) modified the extent of mesial resection with ECoG, took into account preoperative memory disturbance and found more memory decline in patients who were not seizure free to be correlated with lateral extent of resection, but not with a mesial extent of resection. McKhann et al. (2000) used intraoperative ECoG from the HC for tailoring the extent of resection comparing mesial temporal sclerosis (MTS) patients with non-MTS patients and comparing groups with more than 2.5 or less than 2 cm HC resection. They found no correlation between the size of the hippocampal resection and seizure control in the group as a whole, also if stratified by a pathological subtype. Two studies (Cascino et al., 1995; Leijten et al., 2005) used ECoG to tailor the lateral extent of resection. Leijten et al. (2005) resected more lateral neocortex if spikes were present and found a better performance in IQ and visual naming. Cascino et al. (1995) found that the extent of the lateral temporal cortical resection and the ECoG findings are not important determinants of surgical outcome in patients with nonlesional TLE. Jennum et al. (1993) formed two groups determining the extent of lateral lobe resection into those with more or less than 5 cm as guided by ECoG and subdural grid recordings. There was no difference in seizure outcome. Paolicchi et al. (2000) also tailored the posterior extent of lateral resection with ECoG and did not find posterior extent to be of importance but the completeness of the resection of the interictal spike. Kanner et al. (1995) used ECoG and ended up with a wide variety of resection types (six only neocortical resection, nine only amygdala resection, eight amygdala plus hippocampal resection) and had class I outcome in 88% of these mixed groups, concluding from these totally physiologically guided resections that the sparing of HC, or HC and AM or limited resection is not necessarily related to a poor seizure outcome. In all, only one of 10 studies with functional tailoring (mostly by ECoG) found that extent of mesial resection affects outcome.

Retrospective predictor studies

McIntosh et al. (2001) reviewed 126 articles testing 63 factors and identified four major relevant factors associated with good outcome (McIntosh et al., 2001). Extent of mesial resection was not among them and extent of lateral resection could be ruled out to be associated with outcome. In that review six studies were cited (Leonard, 1991; Son et al., 1994; Jooma et al., 1995; Kanner et al., 1995; Salanova et al., 1996; Berg et al., 1998) demonstrating no effect for extent of mesial resection and two studies showing that a greater extent of mesial resection had a significant association with good outcome (Nayel et al., 1991; Wyler et al., 1995). This section focuses only on articles comparing seizure free with nonseizure free patients or patients with good outcome versus those with poorer outcome, if the relation to extent of resection or different types of resection was addressed (Trenerry et al., 1993; Salanova et al., 1996; Berg et al., 1998; Abosch et al., 2002; Clusmann et al., 2002; Bonilha et al., 2004; Joo et al., 2005). Some of these studies mixed temporal and extratemporal cases (Paolicchi et al., 2000), being not so useful for this type of analysis. The study by Clusmann et al. (2002) is a typical retrospective study looking for predictors of good outcome and is of interest because it compared 83 TLRs with 126 SAHs. Their conclusion was that no significant outcome differences were found regarding different resection types but limited resections revealed better neuropsychological results compared to standard TLR. Abosch et al. (2002) conducted a retrospective study of 27 cases with SAH and compared patients with good versus those with poorer outcome. Their main finding concerning suboptimal seizure control was not insufficient resection of HC but that they did not meet the criteria for unilateral MTLE. Consequently they found that 9 of 12 patients with poor outcome did not profit from a repeat operation despite the fact that they had had near total resection of mesial structures. Joo et al. (2005) compared seizure free with nonseizure free patients and defined a larger hippocampal resection predicting better seizure outcome. A similar retrospective predictor study was performed by Trenerry et al. (1993), and they concluded that the resection of nonatrophic HC was associated with a poorer verbal memory outcome. In a retrospective study comparing seizure free with nonseizure free patients and with patients becoming seizure free later Salanova et al. (1996) did not mention extent of resection, similar to Berg et al. (1998) and Janszky et al. (2005) who looked for predictors of good outcome and did not mention extent of resection. In another later comparison of patients with good outcome versus those with unsatisfactory outcome Salanova et al. (2005) found that 86% of the latter group showed residual posterior mesial structures on postoperative MRI.

Postoperative volumetric results

More recent studies have used postoperative MRI to assess extent of resection (Jack et al., 1988; Awad et al., 1989; Siegel et al., 1990; Nayel et al., 1991; Keogan et al., 1992; Cascino et al., 1995; Renowden et al., 1995; Arruda et al., 1996; Jones-Gotman et al., 1997; Vajkoczy et al., 1998; Hadar et al., 2001; Abosch et al., 2002; Bonilha et al., 2004; Salanova et al., 2005; Bonilha et al., 2007; Helmstaedter et al., 2008). Of the authors analyzing their results with regard to extent of resection by MRI technology several independently came to a conclusion similar to “volumetric analysis of MR after SAH indicate that the amount of resected tissues varies considerably” (Novak et al. 2002). This was also underlined by similar statements by Jones-Gotman et al. (1997) and Nayel et al. (1991). The few studies that contain precise data on the variability of resected volumes underlined this with precise numbers (Jack et al., 1988; Siegel et al., 1990; Vajkoczy et al., 1998; McKhann et al., 2000) (Table 4). Jack et al. (1988) described a wide range of resection length (2.5 to 22 mm for parahippocampus in lobectomies and from 10 to 22.5 mm for hippocampi in mesial resection types). In Siegel et al.'s (1990) review of the cases operated by Yasargil the resected tissue varied between 2.1 and 17.7 cm3. In the review of Vajkoczy et al. (1998) of cisternal selective AH, resection varied between 5.7% and 23.6% of temporal lobe volume. This is the variability found in a so-called standardized resection type—selective amygdalohipoocampectomy. In another study (McKhann et al., 2000) the resected mean length based on ECoG tailoring was 28.5 mm for class I and 26.8 mm for combined class II and III outcomes. Overall length of resection varied between 5 and 46 mm. So also in electrophysiologically tailored resection similar group outcomes are achieved with quite variable degrees of resection.

Table 4.  Studies giving volumetric measurements on extent of resection
Vajkoczy et al. (1998, SAH)5.7%–23.6% of temporal lobe volume
Siegel et al. (1990, SAH)2.1 cm3–17.7 cm3
Jack et al. (1988, TLR)2.5–22 mm for parahippocampus
10–22.5 mm for hippocampus
Awad et al. (1989, TLR)1.5–4.5 of 5 possible Index points for basal compartments
0.5–3.5 of 5 possible Index points for mesial compartments
McKhann et al. (2000, EcoG-guided TLR)5–46 mm, mean 28.5 mm Engel I mean 26.8 mm Engel II + III
Jones-Gotman et al. (1997, three resection types)18 of 24 “neocortical” resections reached mesial structures
Awad et al. (1989, TLR)Only 7 of 34 TLR reached mesial structures

In two studies (Awad et al., 1989; Jones-Gotman et al., 1997) postoperative MRI demonstrated that even the type of resection needs verification. Awad et al. (1989) stated that intraoperative estimates and measures routinely overestimate extent of resection. In only 7 of 34 lobectomies in that series the resected compartments included the HC. Jones-Gotman et al. (1997) comparing three resection types from three different centers found that 18 of 24 resections thought to be confined to the lateral lobe in fact included parts of HC or AM.


It is difficult to assess the efficacy of certain resection types in older papers on TLE from times when a clear distinction between MTLE and NTLE was certainly not often made and examined groups most likely consisted of both types of TLE. It is also difficult to directly compare studies from the pre-MRI period with those from the post MRI-period. Even if we let aside those types of study where a tailoring of the lateral or mesial extent of resection was done by performing intraoperative ECoG or by the extent of TL-atrophy, it still remains problematic to form a solidly based opinion on what would be the ideal resection type. Not only are most studies retrospective, many lack postoperative MRI volumetry. Studies with the ideal study design (prospective, randomized) are rare (Table 5), except for a recent study proving the value of operative treatment for drug-resistant TLE (Wiebe et al., 2001). The variability in details of compared resection types does not allow compilation of data in the style of a meta-analysis. The same reservations are true when looking at extent of resection within resection types (SAH or TLR). It remains equally difficult to decide on the necessary extent of resection to achieve the ideal combination of maximal seizure freedom with minimal neuropsychological deterioration. From the present day point of view it is surprising that seizure freedom rates around 50% to 60% or “satisfactory outcomes” of over 60% could be achieved when only the AM and lateral neocortex was removed and the HC was left behind. It is safe to assume that in most of these cases from the pre-MRI era no clear distinction between MTLE and NTLE was safely possible.

Table 5.  Prospective and prospective randomized studies
Wyler et al.1995 70RandomizedTLR with moderate and extensive mesial resection, seizure and neuropsychological outcome, no MRI volumetry
Hermann et al.1999a 30RandomizedSeizure outcome and confrontation naming for TLR with and without STG resection
Hadar et al.2001 28RandomizedSeizure outcome and neuropsychological results for SAH and temporal lobectomy
Lutz et al.2004 80RandomizedNeuropsychology for transcortical versus transsylvian SAH
Awad et al.1989 45ProspectiveSeizure outcome in TLR
McKhann et al.2000140ProspectiveMTS with non-MTS patients determining extent of HC resection with ECoG and evaluation of longer and shorter HC resection
Helmstaedter et al.2008 97ProspectiveNeuropsychological results for corticoamygdalohippocampectomy versus SAH, two-center comparison

SAH versus TLR

Comparing TLR and SAH, McKhann et al. (2006) pointed out that “neither SAH nor TLR can be recommended over the other option as a standard or guideline in the surgical management of TLE.” Keeping the limited quality of evidence in mind it appears that the smaller resection type SAH is associated in the majority of reports with a similar seizure outcome to the larger TLR resection. In trying to explain good outcomes for variable removal extents and different resection types (TLR, SAH, and neocortical resections), Jones-Gotman et al. (1997) hypothesized that “resection either from the HC region or … neocortex results in disconnection of one region from the other so that functionally the two types of resection are similar.” So in summary one can conclude with the reservation that most studies have been performed retrospectively that seizure outcome in SAH was found to be the same as in temporal lobectomy. Comparing neuropsychological outcomes for SAH and TLR there is some evidence, i.e., in 11 of 14 studies, for a slightly better neuropsychological outcome for SAH. It should be pointed out, however, that no single group described a worse neuropsychological outcome for patients with SAH. The various advantages described for SAH vary from author to author but have been described by nine groups, (Table 2) and concern mainly memory.

Intended versus achieved extent of resection

To discuss the aspect of intended versus truly achieved extent of resection may sound surprising. However, in fact it appears adequate to look into this question since the ground-breaking study on extent of resection in glioma surgery where Albert et al. (1994) surprisingly described that compared to the only 21% of glioma cases where the neurosurgeon believed to have residual tumor in fact when checked by MRI in 80% of cases a portion of tumoral tissue was left behind inadvertently. This review of epilepsy surgeries has made a few things clear: Although the neurosurgeon thinks he is performing a relatively standardized type of resection, the effect of that resection as measured in cubic millimeters or in millimeter hippocampal length may be quite different. In the few studies where postoperative volumetric MRIs were performed the extent of mesial resection was highly variable (Awad et al., 1989; Siegel et al., 1990; Nayel et al., 1991; Jones-Gotman et al., 1997; Abosch et al., 2002). The aspect of variability not only covers the fact that similar rates of seizure freedom can be achieved with a standard lobectomy versus SAH, and not only the fact that groups with quite different extents of mesial resection achieve similar results, but also the fact that volumetric results obtained after mesial resection showed a variability in extent of resection stretching up to eight times the minimal resection volume for cases with maximal resection (Table 4) (Siegel et al., 1990). Preliminary results from our own ongoing volumetric study also demonstrate a high degree of variability in extent of mesial resection in two groups, one with an intended minimal resection of 25 mm from tip of temporal horn and the other one with an intended “long” resection of more than 35 mm. In a preliminary analysis of 69 cases the true extent of resection from pole posteriorly varied between 40% and 90% for the “short” group and between 50% and 100% of total HC volume for the “long” group. Epilepsy surgeons should take note of these facts, which demonstrate a truly amazing variability. This variability may only in part be explained by the variable degree of hippocampal atrophy. There is certainly a number of other factors influencing these highly variable degrees of resection, such as degree of exposure, difficulty of opening the sylvian fissure, extent of transection of anterior aspect of temporal stem, extent of opening of temporal horn, extent of atrophy with larger or smaller ventricle size, more or less posterior mesial exposure and access depending on angle of head rotation and others.

Several authors have previously pointed out the need to assess the extent of resection precisely (Nayel et al., 1991), underlining the fact that “intraoperative linear measurement … is often imprecise … .” It has been demanded therefore “that evaluation of outcome requires postoperative morphometric review of the surgical lesion” (Novak et al., 2002), or that “extent of resection must be assessed by neuroimaging” (Jones-Gotman et al., 1997). Here it has become clear that a very important differentiation has to be made between the declared and intended extent of resection and the truly achieved extent of resection.

Extent of resection

In this review 12 studies were included with a conclusion concerning the extent of resection. Six authors did not find that a larger extent of resection led to better seizure outcome, five studies found it did. Wyler et al. (1995) made a very strong recommendation for more posteriorly reaching hippocampectomy in his prospective randomized study. It should be pointed out that Tonini et al. (2004) in their review, also considering extratemporal series and tumor cases described that extensive surgical resection belonged to the stronger prognostic indicators of seizure remission. But the same authors stated that “firm conclusions cannot be drawn for extent of resection.” These authors cited 10 reports where extent of resection was reported, however, among the 10 were two which only looked at lesionectomy cases (Li et al., 1997; Morris et al., 1998) or another two included mixed extratemporal and temporal cases (Rossi et al., 1994; Zentner et al., 1996). In a fifth study it was not a smaller or larger extent of resection but the completeness of the resection of the interictal spike zone. The interpretation therefore in at least 5 of the 10 studies cited as demonstrating a positive prognostic value for a larger extent of resection appears questionable with regard to the extent of resection for MTLE. In a recent study (Bonilha et al., 2007) five publications were cited as demonstrating a “positive correlation between the extent of hippocampal removal and the success of surgery” but three of the five studies (Kuzniecky et al., 1993; Hennessy et al., 2000; Salanova et al., 2005) did not measure and correlate extent of resection with outcome. It appears that not only the facts are difficult to sort out but in the discussion the results have not always been interpreted precisely.

Thus, concerning the extent of resection the situation is not quite decided (Fig. 1). Although most authors report that seizure outcome is irrespective of the extent of mesial and lateral resection, five authors describe that seizure outcome is better if larger volumes or specific substructures are removed. However, the retrospective study type does not provide class I evidence.

Figure 1.

Bar graph showing numbers and percentage of studies indicating the major result of studies comparing seizure and neuropsychological outcome in the two resection types SAH and TLR and the relationship between extent of resection to the two outcome parameters. Please note that the antagonistic categorization is somewhat schematic, Tables 1, 2, and 3 describe the main outcome tendency of these studies in more detail and contain authors' names and references.

How to explain the contradictory findings

The reviewed results on the significance of resection extent are obviously quite contradictory, concerning the extent of lateral resection in TLR as well as extent of mesial resection. Despite the very remarkable outcome difference for the shorter and longer resection group in the only randomized study (Wyler et al., 1995) even the extent of mesial resection in the more restricted SAH is still under debate. This is in contrast to the fact that the larger series from experienced centers employing quite different presurgical evaluation techniques and variable resection protocols publish rates of seizure freedom or satisfactory seizure outcome (combined class I and II) in about the same range. If extent of mesial resection is so important according to Wyler's study (1995), why then do different centers have very similar results, including those centers who do not try to go for maximum mesial extent of resection.

If the dogma of maximal mesial resection would be the only explanation, cases without seizure freedom, which had hippocampal reresection should markedly improve. This expectation contrasts with several reports where seizure freedom rates for the reoperation was not 85% or 90% but only 20% (Hennessy et al., 2000), 30% (Abosch et al., 2002) or 50%–60% (Wyler et al., 1989; Awad et al., 1991; Germano et al., 1994; Salanova et al., 2005). As discussed previously (Nayel et al., 1991; Abosch et al., 2002), the explanation could be that lateral neocortical structures are involved in ictogenesis, or that initially bitemporal ictal activity had been present. Different approaches resulting in different extent of postoperative fiber tract damage (Baxendale et al., 2000) or different possibilities for damage to surrounding structures (Helmstaedter et al., 2004) have been discussed. The most interesting explanation, however, to explain these diverging results from studies looking at extent of resection could be that it is not at all that important to completely excise all mesiobasal structures (HC, PHG, and AM) but that it would be enough to resect or interrupt a large enough part of the mesial structures to render the mesiobasal network between HC, PHG, and AM unable to build up and sustain a seizure. Olivier's group (Abosch et al., 2002) called this “severing a critical proportion of the connections” when they discussed the variability in extent of resection described by McKhann et al. (2000).

In conclusion it can be summarized that there is very little class I evidence (i.e., prospective, randomized, and MRI volumetry controlled) favoring a larger extent of resection and that the majority of retrospective studies does not conclude that a larger extent of resection is reliably associated with a better seizure outcome.

What is missing is a prospective randomized trial comparing the two resection types SAH and TLR and a prospective randomized trial comparing a small and a large extent of mesial resection, where both trials should be controlled by postoperative MRI volumetry (Table 6).

Table 6.  Requirement for a trial on significance of resection extent for seizure and neuropsychological outcome after TLE surgery
Prospective randomized design for extent of mesial and lateral resection
Pre- and postoperative MRI volumetric
Identical neuropsychological evaluation pre- and postoperatively
Sufficient follow up of 1 year minimum
Large enough case numbers to allow for subgroup analyses (e.g., R vs. L TLE, MTS vs. non-MTS, pre-op. small neuropsychol. deficits vs. larger pre-op. neuropsychological deficits) dropouts, and nonparticipating patients
An oligocentric study may be necessary to cover above requirements, necessitating the use of similar neuropsychological test batteries and presurgical evaluation strategies


Class I evidence concerning seizure outcome related to type of resection and extent of resection of mesial temporal lobe structures is hardly available. SAH appears to have similar seizure outcome and possibly a better cognitive outcome than TLR. It remains unclear whether a larger mesial resection extent leads to better seizure outcomes. More prospective randomized studies with MRI volumetrics of extent of resection are needed.


This study was supported by a Grant of the Deutsche Forschungsgemeinschaft, DFG (SFB TR3, TP1). The useful comments of T.N. Lehmann, M.D. and Prof. C. Helmstaedter are gratefully acknowledged. E. Heunemann gave excellent secretarial support.

Conflict of interest: I confirm that I have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines. I confirm that there is no conflict of interest.