Neuropsychological Outcome after Selective Amygdalohippocampectomy with Transsylvian versus Transcortical Approach: A Randomized Prospective Clinical Trial of Surgery for Temporal Lobe Epilepsy


Address correspondence and reprint requests to M.T. Lutz at Department of Epileptology, University of Bonn, Neuropsychology, Sigmund-Freud-Strasse 25, 53105 Bonn, Germany. E-mail:


Summary: Purpose: Selective amygdalohippocampectomy (SAH) is a surgical treatment option for patients with medically intractable mesial temporal lobe epilepsy. In contrast to standard anterior temporal lobectomy, resection of unaffected tissue is limited, although it achieves equal seizure outcomes in selected patients. In SAH, the mesial structures can be approached by different routes, the transsylvian approach and the transcortical approach. Advantages or disadvantages with respect to postoperative cognitive outcome are still a matter of debate.

Methods: Eighty randomized patients were included in the analyses. In 41 patients, the transsylvian approach, and in 39 patients, the transcortical approach was performed. All patients received comprehensive neuropsychological testing of verbal and nonverbal memory, attention, and executive functions before and 6 months or 1 year after SAH.

Results: Seventy-five percent of patients became completely seizure free with no difference depending on the chosen approach. Repeated measures multivariate analysis of variance (MANOVA) showed that cognitive outcomes after both approaches were essentially the same. The only exception was phonemic fluency, which was significantly improved after transcortical but not after transsylvian SAH.

Conclusions: The results indicate that either surgical approach can be chosen independent of cognitive outcome criteria. Improvement in phonemic fluency after transcortical SAH may reflect selective normalization of cognitive function after epilepsy surgery, whereas frontal lobe manipulation might have hindered recovery of this function after transsylvian SAH.

Randomized prospective trials of surgery for mesial temporal lobe epilepsy (MTLE) have virtually not been conducted. Because of this lack of data, it is difficult to decide which surgical techniques should be preferred. However, the relative high prevalence rate of mesial temporal sclerosis (MTS) as a cause of intractable epilepsy (1,2) stresses the high importance of conducting randomized studies.

A treatment option for medically intractable MTLE is the surgical resection of the amygdale–hippocampus complex. The standard operation in many centers is the anterior temporal lobectomy (ATL), in which the anterior part of the temporal lobe, including the mesiobasal temporal structures, is resected.

In the belief that potentially functional temporal neocortex in cases with a clearly demonstrated mesiobasal seizure focus should be spared to prevent from postoperative neuropsychological sequelae, the selective amygdalohippocampectomy (SAH) was suggested as a more limited surgical option. The short-term and long-term cognitive benefits of SAH compared with ATL have been shown by different groups (1,3–6); however, the results are equivocal (7).

In SAH, the epileptogenic foci of the MTL can be approached by different operative routes. Two frequently used approaches are the transsylvian approach through the deep sylvian fissure (8) and the transcortical approach (e.g., through the middle temporal gyrus) (9,10). Both approaches are associated with advantages and limitations but have been shown to result in similar favorable seizure outcome (11,12).

Although to a smaller degree, even SAH carries the risk of significant brain damage in addition to the intended excision of the mesiobasal structures. Secondary brain damage may be caused by collateral damage following the surgical routes for the access to the deep structures, wallerian, or transneuronal degeneration. Each of these factors may appear differently after following different surgical approaches.

In the transsylvian approach, unnecessary resection of nonepileptogenic temporal neocortex is largely avoided. However, ∼20% of the anterior temporal stem of the superior temporal gyrus must be transsected (6). In addition, transsylvian SAH carries the risk of vascular injury or vasospasm, attributed to the close proximity of the dissection to major vascular structures in the sylvian fissure (13,14). In contrast, during the transcortical approach, a portion of the lateral temporal neocortex must be dissected to provide access to the ventricle and the mesiotemporal structures (10). A recent study from our center confirmed that both approaches cause “collateral damage” (15).

Due to either damage of normal tissue or the interruption of neuronal circuits (7,16), cognitive morbidity can occur after SAH. Cognitive outcome is, apart from freedom from seizures, one of the most important criteria of surgical success. Preservation or improvement of cognitive functioning is directly related to the quality of life of operated-on patients. The most consistent report of cognitive decline is decline in verbal memory after left-sided SAH (17).

Improvement after SAH also may occur. An increase of the metabolic activity has been reported in the contralateral hippocampus and the orbitofrontal cortex after transsylvian SAH (18). With regard to cognitive functioning, improvements may occur on a group level as well as on an individual level. One possible reason for short-term and long-term release of cognitive functions after surgery is the successful control of seizures and interictal epileptogenic discharges (4). Even asymptomatic discharges suppress cognitive function of the temporal lobe (19). Postoperative release-of-function phenomena have been explained by a reduction of the propagation of abnormal discharges from the epileptogenic to healthy brain regions [relating to ATL; see (20,21)].

The aim of the present study was the comparison of the transsylvian and the transcortical approaches with respect to neuropsychological outcome. To allow meaningful conclusions, stringent methodologic criteria are required. To allow comparisons between types of approach as well as sides of surgery, investigations of sufficient sample sizes are necessary. Use of parallel versions of the tests reduces retest biases, and—most important—randomization should be performed. Furthermore, to explore effects of lesions in adjacent and remote areas in relation to the operation area, one should not solely rely on memory measures but also use a comprehensive test battery to assess different cognitive domains. Considering these methodologic issues, the results of the study will be of practical and theoretical interest.

To our knowledge, no prospective study compares cognitive outcome after following different approaches for SAH. Preliminary reports of this project (15,22) incorporated only subpopulations and focused on different issues.



The original sample consisted of a consecutive series of 140 patients who had been randomly assigned to SAH with either the transsylvian or the transcortical approach for treatment of intractable seizures of MTL origin and the magnetic resonance imaging (MRI) diagnosis of MTS. All patients gave informed consent for randomization and participation in this study. The study was approved by the ethics committee of the University Bonn medical center (ref. 229/00).

Patients had to meet the following criteria: (a) presence of hippocampal sclerosis or gliosis on MRI, (b) age 16 years or older, (c) absence of retardation (IQ >69), and (d) availability of complete preoperative and postoperative neuropsychological evaluations. Moreover, left-handed patients as assessed with the Oldfield Inventory (23), as well as patients with atypical speech representation demonstrated by intracarotid amobarbital testing or functional MRI, were excluded in the present sample. Two patients with postoperative neurologic complications (hematoma) had to be excluded. Twenty-seven patients had to be excluded because of missing data sets. From the remaining patients, 23 had atypical speech representation or were not right-handed. A subset of 88 patients resulted. To ensure the reliability of statistical procedures (see later) that afford a well-balanced design, eight successive patients of the largest two subgroups (the right and the left transsylvian groups) were randomly selected and excluded. Finally, 80 patients were included in this study. Forty-one patients were operated on via the transsylvian approach, and 39 patients were operated on via the transcortical approach. In the transsylvian group, 22 patients showed right-sided AHS, and 19, left-sided AHS; in the transcortical group, right-sided AHS was shown in 18 patients, and left-sided AHS, in 21 patients.

The age of patients ranged from 16 to 60 years (mean, 36.76; SD, 9.72) and had between 7 and 18 years of education (mean, 10.03; SD, 2.33). Half of the patients were women. Verbal intelligence was evaluated in a subgroup of 68 patients with a vocabulary estimate (24), similar to the widely used New Adult Reading Test (NART). Verbal IQ ranged from 76 to 143 (mean, 101.69; SD, 13.99). Table 1 provides group demographic characteristics and seizure history information.

Table 1. Patient characteristics
Group Side of surgeryTranssylvian SAHTranscortical SAHGroup difference
Right (n = 22)Left (n = 19)Right (n = 18)Left (n = 21)
  1. The cells contain the number of patients for gender, and group means (standard deviation in parentheses) for the other variables.

  2. SAH, selective amygdalohippocampectomy; AEDs, antiepileptic drugs; NS, not significant (p > 0.05).

  3. aχ2 test.

  4. bAnalysis of variance (ANOVA).

  5. cMann–Whitney U test.

  6. dFisher's exact test.

  7. en = 68.

Age at surgery(yr)36.36 (11.99)36.58 (8.49) 40.72 (9.14)33.95 (7.98) NSb
Age at onset of epilepsy (yr)10.23 (7.87) 10.05 (10.38) 13.67 (10.40)11.00 (10.78)NSb
Duration (yr)26.14 (11.78)26.53 (13.11) 27.06 (13.06)22.95 (11.21)NSb
Retest interval (months)7.36 (5.44)7.54 (5.12) 6.28 (4.95)7.92 (4.59)NSb
No. of AEDs2.10 (0.89)2.28 (1.13) 1.88 (0.99)1.75 (0.55)NSc
No. of patients with intracranial recordings6454NSd
Education(yr)10.00 (2.66) 9.84 (1.50)11.06 (3.49)9.33 (0.73)NSb
Verbal IQestimatede104.47 (15.54) 97.39 (13.32)111.86 (14.96)95.79 (6.49) Side (L < R)
 p < 0.001b

Surgical groups did not differ significantly in terms of age, education, or gender distribution. No differences were found according age at onset of epilepsy, duration of epilepsy, age at surgery, and numbers of antiepileptic drugs (AEDs). The verbal IQs of the group with left TLE were significant lower than the IQs of the right TLE group, but no difference in verbal IQs was found between the surgical groups (transsylvian vs. transcortical).


Preoperative investigations

Extensive preoperative diagnostics demonstrated a unilateral temporomesial seizure onset in all cases and yielded the diagnosis of MTLE. The details of presurgical epileptologic diagnostics at our center have been described previously (25). Site of suspected seizure onset was identified presurgically in the patients by means of high-resolution MRI, ictal and interictal scalp EEG videotelemetry monitoring, ictal clinical semiology, and neuropsychological examination. Additionally, in a subgroup of 19 patients, intracranial recordings were applied. No significant differences were noted between surgical groups according to frequency of intracranial recordings (Table 1).

Surgical procedures

Surgeries were performed between 1999 and 2001 at the Department of Neurosurgery at the University of Bonn, Germany. Patients were assigned randomly to the transsylvian or to the transcortical approach.

In brief, the transsylvian approach was performed via a pterional craniotomy of ∼5 cm in diameter. After microsurgical dissection of the sylvian fissure (2.5–3 cm), the temporal horn of the lateral ventricle is entered through the temporal stem, thus allowing resection of the mesial structures (amygdala, hippocampus, and parahippocampus). The collateral sulcus is the intended lateral resection border, and the area of maximal brainstem diameter is the dorsal border.

In the transcortical keyhole approach, a 3-cm craniotomy is centered on the projection of the middle temporal gyrus. Neuronavigation was used to determine the optimal placement of a 2-cm corticotomy in the middle temporal gyrus and to direct the white matter cleavage to the temporal horn of the lateral ventricle. Once inside the ventricle, we resect the mesial structures, as described earlier.

In 76 cases, hippocampal sclerosis was proven histologically, and in three patients, hippocampal gliosis was found. No pathological features were found in one patient.

Neuropsychological assessments

Neuropsychological assessments typically took place within a few months before surgery and at an average of 7.3 months after surgery. All patients were maintained with stable AED therapy at the time of the two evaluations. The neuropsychologists were blinded to the type of surgical resection performed. Several standard neuropsychological instruments were administered. Measures included tests of verbal and nonverbal memory, attention, and executive function. If appropriate, parallel versions of the tests were used in the postoperative evaluation.

Verbal and nonverbal memory were assessed with two serial learning tasks. The verbal learning task is a German version of the Rey Auditory-Verbal Learning Test (26). The subject has to learn a list of 15 nouns in five consecutive trials, each trial followed by a free-recall test. After presentation of an interference list of 15 words, the first list has to be recalled again immediately and after a 30-minute delay period. Finally, recognition of the first 15 words was tested. Parameters chosen for statistical analysis were learning capacity (correct reproduced words in trial one to five), loss in free recall after 30-min delay, and recognition performance (adjusted for mistakes). Nonverbal memory was assessed by using a design list learning test, which requires learning and reconstruction of nine abstract designs in five consecutive learning trials. Each design consists of five lines of the same length. Recall requires reconstruction of the designs by five sticks of equal length. After a 30-min delay, recognition of the designs from alternatives was tested (27,28). Parameters chosen were learning capacity (correctly reconstructed figures in five consecutive trials) and recognition performance (adjusted for errors).

Attention was assessed in speed and quantity. The processing speed was tested by a symbol-counting task [subtest 1 of the c.I. test (29)] and by the time score of a maze test (30). In the latter task, the patient had to trace three mazes of increasing difficulty with a pencil.

Short-term capacity was assess with two memory-span tasks: the Digits Forward test [HAWIE-R, German adaptation of the WAIS (31)], and the Tapping Forward test [modified adaptation of Corsi block test after Milner (32,33)].

Five tests were conducted to examine executive functions: Verbal fluency was assessed by using a written word-fluency test, a subtest of a German test of intelligence [Leistungsprüfsystem (34)]. This task requires the subject to write down as many words as possible in 60 s that begin with a designated letter (three trials). For assessment of working memory, the subject must repeat digits in reverse order [Digit Backward task of HAWIE-R (31)]. The error score of the maze test (30) was used to provide data about foresight/anticipation and feedback-guided decision making. The second subtest of the c.I. test (29) assesses response inhibition or interference effects, respectively. It requires inverse reading of two rows of “A” and “B” (AABAB as BBABA). Time needed to perform the tasks was the object of the evaluation. Finally, a motor sequencing task adopted form Luria (35) was conducted to elicit various deficits of sequential motor organization. The task requires rapid alternation of both unimanual and bimanual motor sequences.

Seizure outcome

Patients were categorized as seizure free if they were completely free of seizures with no auras in the interval between surgery and the postoperative neuropsychological evaluation [class 1 by using the proposal for a new classification of outcome of the International League Against Epilepsy (36)].

Statistical analyses

Preoperative group differences in cognitive functioning were assessed by multivariate analysis of variance (MANOVA) specifying surgical approach (transsylvian vs. transcortical) and laterality (left vs. right) as group factors. Separate MANOVAs were performed for each cognitive domain (verbal memory, nonverbal memory, attention, and executive function).

Postoperative changes were assessed by repeated measures MANOVA with approach (transsylvian vs. transcortical) and laterality (left vs. right) as between-group factors and test repetition (preoperative vs. postoperative) as the repeated measure or within-group factor. Again, separate MANOVAs were performed for each block of dependent variables (verbal and nonverbal memory, attention, executive function). In the presence of a significant omnibus MANOVA F, separate univariate analyses of variance (ANOVAs) for each dependent variable per block were performed to determine the locus of the significant difference. Changes within groups were assessed with post hoc t tests for paired samples. According to the aim of this study, the interaction between approach and test repetition was of greatest interest.


Seizure outcome

Overall, 60 (75%) patients were classified as completely seizure free ∼7 months after surgery. No significant association was found between surgical approach and seizure outcome (χ2= 0.150, df= 1, p = 0.80). Thirty (76.9%) of the 39 transcortical SAH patients and 30 (73.2%) of the 41 transsylvian SAH patients were free from seizures.

Preoperative group differences

The neuropsychological test results of the subjects are summarized in Table 2A and B. Two-factor MANOVAs with approach and laterality as between-group factors showed a significant main effect of the side of seizure onset on verbal memory [F(3, 73) = 2.99, p = 0.036]. Univariate analyses indicated a significant difference in verbal learning [F(1, 75) = 3.99, p = 0.05] and in verbal delayed recall [F(1, 75) = 7.85, p = 0.006] with the right MTLE group outperforming the left MTLE group in both measures. No significant preoperative group differences were found for nonverbal memory, attention, or executive function.

Table 2A. Neuropsychological test results, transsylvian SAH
 ParametersRight-sided SAHLeft-sided SAH
  1. Results are presented as raw scores. The cells contain the means (standard deviation in parentheses).

Verbal memory
 LearningMaximum 75 words46.14 (9.71) 44.64 (11.33)41.63 (12.59)36.11 (10.91)
 Loss in free recallNo. of words lost2.64 (2.85)3.27 (2.95)3.47 (2.34)4.47 (1.98)
 RecognitionMaximum 15 words10.50 (4.06) 9.86 (4.38)8.84 (5.44)4.26 (6.01)
Nonverbal memory
 LearningMaximum 45 designs13.73 (9.00) 12.95 (11.29)14.74 (8.45) 17.00 (11.33)
 RecognitionMaximum 9 designs4.45 (3.00)3.95 (3.11)4.00 (2.60)4.95 (4.25)
 Maze testTime (s)347.27 (157.81)358.82 (224.51)322.39 (117.69)341.67 (207.37)
 Memory spanVerbal span plus Nonverbal span10.91 (1.80) 11.50 (2.11) 10.67 (1.61) 11.61 (2.09) 
 Symbol countingtime (s)18.36 (6.40) 17.05 (5.64) 17.33 (3.97) 18.89 (8.41) 
Executive function
 InterferenceTime (s)23.55 (7.99) 22.86 (8.91) 26.72 (8.05) 23.72 (6.05) 
 Maze testTotal error score11.09 (6.57) 9.27 (5.76)14.22 (7.64) 9.28 (5.09)
 Verbal fluencyNo. of words across 3 min29.32 (11.58)28.77 (11.07)25.22 (9.30) 23.67 (8.25) 
 Working memoryLongest sequence backwards, verbal4.32 (1.21)4.55 (1.22)3.83 (1.25)4.39 (1.24)
 Motor sequencesTotal error score6.64 (3.23)6.64 (3.16)6.39 (3.29)6.44 (2.55)
Table 2B. Neuropsychological test results, transcortical SAH
 ParametersRight-sided SAHLeft-sided SAH
  1. Results are presented as raw scores. The cells contain the means (standard deviation in parentheses).

Verbal memory
 LearningMaximum 75 words48.00 (9.55) 47.71 (5.72) 43.14 (9.41) 34.24 (9.94) 
 Loss in free recallNo. of words lost2.00 (1.80)3.29 (1.79)4.10 (2.00)4.38 (1.50)
 RecognitionMaximum 15 words10.65 (3.67) 10.47 (5.21) 8.90 (5.25)2.38 (7.84)
Nonverbal memory
 LearningMaximum 45 designs12.53 (9.70) 14.67 (8.57) 14.67 (10.14)15.78 (11.62)
 RecognitionMaximum 9 designs4.07 (3.49)5.27 (2.76)4.39 (3.68)4.83 (4.38)
 Maze testTime (s)405.56 (123.77)328.31 (168.03)324.95 (192.49)305.05 (187.49)
 Memory spanVerbal span plus nonverbal span11.37 (2.03) 12.19 (1.68) 10.95 (2.27) 11.16 (2.17) 
 Symbol countingTime (s)18.25 (5.66) 18.19 (5.28) 16.68 (5.74) 17.21 (4.78) 
Executive function
 InterferenceTime (s)21.73 (4.38) 21.73 (4.83) 22.21 (10.46)22.05 (6.00) 
 Maze testTotal error score7.00 (5.61)7.80 (6.75)13.79 (8.55) 9.84 (7.21)
 Verbal fluencyNo. of words across 3 min30.20 (9.33) 35.27 (11.35)26.58 (10.34)28.89 (11.49)
 Working memoryLongest sequence backwards, verbal4.73 (1.10)4.87 (1.41)4.21 (1.36)4.68 (1.97)
 Motor sequencesTotal error score5.87 (2.39)5.53 (1.96)6.95 (3.81)6.26 (3.18)

Postoperative change in neuropsychological performance

MANOVAs for each cognitive domain (verbal memory, nonverbal memory, attention, and executive function) showed the following results.

Verbal memory

Concerning verbal memory, main effects of test repetition [F(3, 73) = 7.725, p < 0.001] and side of surgery [F(3, 73) = 7.454, p < 0.001] were found. Univariate analysis indicated a significant loss after surgery in all three parameters of verbal memory [all F values (1, 75) ≥6.05 with p ≤ 0.016]. The left MTLE group performed worse than the right MTLE group in all three parameters [all F values (1, 75) ≥11.414 with p ≤ 0.001]. Furthermore, the interaction between test repetition and side [F(3,73) = 6.295, p = 0.001] reached significance, basically caused by the recognition and learning parameters [F(1, 75) ≥8.841, p ≤ 0.004]. T tests for paired samples indicated a significant decline in recognition performance (t= 4.91; p < 0.001) and in learning performance (t= 4.34, p < 0.001) for the left TLE group after surgery. Taken together, decline in verbal memory postoperatively was mainly found after left-sided surgery.


In attention, MANOVA yielded, independent of approach and side of surgery, a main effect of test repetition [F(3,69) = 3.149, p = 0.030] because of an improvement in the memory span tasks [F(1, 71) = 8.274, p = 0.005].

Executive function

Concerning executive functions, a main effect of test repetition was found [F(5, 66) = 3.871, p = 0.004], reflected by an improvement in the working memory task [F(1, 70) = 4.766, p = 0.032] and in the error score of the maze task [F(1, 70) = 17.037, p < 0.001]. Furthermore, the interaction between test repetition and side of surgery reached significance [F(5, 66) = 2.864, p = 0.021], basically because of the error score of the maze task [F(1, 70) = 10.755, p = 0.002]. Paired comparisons with t tests showed that a strong improvement in this score appeared exclusively after left-sided surgery (t= 5.00; p < 0.001). No change in performance was seen after right-sided surgery (t= 0.94; p = 0.35).

Interaction between test repetition and approach

Of greater interest was the interaction between test repetition and approach, which reached significance only in executive function [F(5, 66) = 3.310, p = 0.010]. Post hoc univariate ANOVAs showed significant group differences in verbal fluency performance [F(1, 70) = 9.644, p = 0.003]. T tests for paired samples indicated a gain in verbal fluency after transcortical SAH (t =–2.754; p = 0.009), see Fig. 1.

Figure 1.

Interaction between test repetition and operative approach. The word fluency improved postoperatively only in patients with transcortical selective amygdalohippocampectomy (SAH). In patients with transsylvian SAH, preoperative and postoperative word fluency were similar.

No correlation was found between improvement in word fluency and postoperative freedom from seizures [χ2 = 0.312 (df= 1), p = 0.404; Fisher's exact test]. An analysis of covariance with the reduction of AEDs and the verbal IQ added as covariates did not change the significant interaction between test repetition and approach in word fluency [F(1, 54) = 6.206, p = 0.016].

Additional analyses: RCIs

Additional analyses were undertaken to confirm the group-mean comparisons by indication of real postoperative changes on individual patients. For this purpose, reliable change indices (RCIs) were calculated to confirm changes in verbal fluency on an individual basis (37). RCIs were based on test–retest data of an independent sample of 62 epilepsy patients. Patients were referred to our center for presurgical evaluation or for optimizing AED therapy. No intermediate surgical treatment took place. The mean test–retest interval was 11.6 months (SD, 6.2 months), and the mean age was 33 years (SD, 13 years). To determine the effect of epilepsy surgery independent of test–retest artefacts like practice effects, the 90% confidence interval (CI) was calculated under consideration of the retest bias. According to this procedure, an improvement of nine or more words in the phonemic word fluency task can be considered a reliable improvement. In the total sample of this study, reliable improvement in word fluency according to RCI was found in 13 (16.9%) patients. In the transcortical group, an improvement was noted in 11 (29.7%) of 37 subjects; in the transsylvian group, only two (5%) or 40 subjects showed an improvement >8 points (χ2 = 8.377; p = 0.004). To evaluate whether these proportions are statistically significant, we calculated binominal tests, assuming—based on the 90% CI—that a change in either direction should occur by chance in 5% of the cases (17). The binominal tests indicated that the gain in word fluency after surgery with the transcortical approach was significantly higher than expected (p < 0.001), but not after surgery with the transsylvian approach (p = 0.642).


The present study was designed to assess differences in cognitive outcome after two types of SAH, the transsylvian and the transcortical approach for treatment of medically intractable MTLE. One important feature of this study is the use of a comparative test battery that incorporates measures of mesial and cortical temporal functions (e.g., learning and memory), as well as measures of other cognitive domains like attention and executive functions. Moreover, all surgeries were performed at one center. Probably the most important feature of the present study is the randomized assignment to the surgical approaches. This unpredictability in the treatment assignment holds true even if the assignment was predetermined because of limited availability of neuronavigation in five patients. By following this procedure, most of the typical shortcomings of comparative studies are resolved, and no time, selection, or center bias has to be suspected. These features enable us to compare test results of the two surgical groups directly.

In terms of freedom from seizures, both approaches resulted in virtually the same outcome. This is in line with other studies comparing different series of temporal lobe resections for MTLE (e.g., 1,38). Abosch et al. (16) stated that in view of the often comparable reported freedom of seizures, various techniques all result in severing a critical proportion of the connections between the entorhinal cortex, hippocampus, and fimbria–fornix. The same may be true in regard of the transsylvian and transcortical approaches.

Taking into consideration that each test battery has its blind spots, the most surprising result of this study is the nearly complete lack of clear differences in cognitive outcome after both approaches. This refers to both the postoperative improvements and the postoperative losses in performance. At first, the question whether this lack of differences could be an artefact caused by insufficient statistical power should be addressed. For this purpose, we conducted a compromise power analysis by using the program G*Power (39,40). Parameters specified were the sample size, an estimated correlation between the repeated measures of r= 0.50, and an expected “medium” effect according to Cohen's effect-size conventions (41). The ratio q = beta/alpha, which specifies the relative seriousness of both errors, was set at unity. This means that both types of errors were considered equally serious. The power was calculated for the univariate interaction between test repetition and surgical approach. The estimated power for the interaction term was 0.98 (with alpha = 0.02). After this, the power to detect an expected “medium” effect can be seen as reasonably high. This high power may be attributable partly to the high efficiency of a study design with dependent measures.

Performances in verbal memory are not only of theoretical interest, because they determine performance in every day activities as well as subjective well-being of the patients (42). Different from our expectation resulting from comparisons of left anterior 2/3 resections and SAH, sparing the lateral cortical tissues did not result in an advantage of the transsylvian approach. Left-sided surgeries thus led to decline in verbal memory independent of the chosen approach. This finding is fully in line with the finding resulting from a small subsample of patients with a transsylvian or transcortical approach in whom postoperative memory decline was associated with the extension of the collateral cortical damage rather than with the type of approach (15). Another argument could be that in both groups, preexisting damage of the lateral part of the neocortex due to long-lasting MTLE may prevent significant gains by sparing lateral tissues (7,43). This would be in line with studies reporting a lack of differences in verbal memory between SAH and ATL (7). Therefore one could speculate that the postoperative deterioration of verbal learning and memory results from the disconnection of memory circuits by resection of the hippocampus after SAH in general, and residual memory capabilities of the lateral temporal cortex can be superimposed by this hippocampal transsection.

With the exception of verbal fluency changes, other cognitive domains also proved to be independent of the type of approach. The nonverbal memory performance was unchanged after surgery, and improvements were found in the memory span and in the working memory task. Regarding foresight/anticipation assessed by the maze test, a selective improvement was found after left-sided surgeries.

Considerable improvements in verbal fluency were evident particularly after the transcortical approach. This strong effect cannot be attributed to random factors and needs further explanation. Improvements in verbal fluency after epilepsy surgery is a common phenomenon that has been attributed to achieved freedom from seizures (21). Because no differences in achieved seizure freedom were seen between the groups, the present results do not support the “nociferous cortex” hypothesis in a general form, that freedom of seizures may result in a release of previously suppressed functions or capacities for compensation (20).

Verbal fluency tasks are typically interpreted as an executive function, mediated by the frontal lobe cortex. Improvements in verbal fluency may reflect selective normalization of frontal lobe function after epilepsy surgery. In the case of transsylvian SAH, however, frontal lobe manipulation during surgery might have hindered the expected process of recovery. A recent report of our group shows that cognitive effects after surgery are partially related to the so-called “collateral damage,” which may extend to the frontal lobe in case of transsylvian approaches (15). Thus the observed changes of verbal fluency performance could be a result of simultaneous or interactive effects of recovery and consequences of new lesions or disturbances after surgery. It cannot be excluded that the differences in the verbal fluency performance disappear with longer postoperative intervals. A post hoc comparison between transsylvian patients with early versus late postoperative examinations did not show a difference in the extent of change in verbal fluency. However, valid differentiation between possible processes of recovery and relatively fixed effects of new lesions can be attained only by longitudinal studies with more than one postoperative assessment.

Bearing in mind that reliable differences between the approaches showed up in just one subtest of a comprehensive test battery, the transcortical approach seems to be only slightly preferable from a neuropsychological point of view. This should contribute to decision finding, but we are aware that a variety of other surgical and epileptologic aspects determine the choice of the approach for surgical treatment of medically intractable MTLE.