Memory Outcome after Selective Amygdalohippocampectomy: A Study in 140 Patients with Temporal Lobe Epilepsy


Address correspondence and reprint requests to Dr. U. Gleissner at University Hospital of Epileptology, Sigmund-Freud Str. 25, 53105 Bonn, Germany. E-mail:


Summary:  Purpose: The technique of selective amygdalohippocampectomy (SAH) was originally developed in epilepsy surgery to spare unaffected brain tissue from surgery, thus minimizing the cognitive consequences of temporal lobe surgery. The results of previous studies, however, are equivocal in this regard. This study evaluated memory after SAH in a large sample of patients with mesial temporal lobe epilepsy.

Methods: The 140 patients received material-specific memory tests before and 3 months after unilateral SAH.

Results: Significant declines in all aspects of verbal learning and memory were found particularly for the left resected group. With reliability-of-change indices, a high number of patients showed postoperative verbal memory declines, ≤51% in left SAH and ≤32% in right SAH. For left SAH, a higher preoperative verbal memory performance, a lower preoperative nonverbal memory score, an older age at surgery, and a later onset of epilepsy predicted a stronger decline in verbal memory. After right SAH, the risk for a verbal memory decline was slightly increased when patients had surgical complications or a presurgical evaluation with bilateral intrahippocampal depth electrodes. Results concerning nonverbal memory were less clear.

Conclusions: The results clearly indicate, that particularly left SAH can lead to a significant decline in memory functions. Predictors of postoperative verbal memory were similar to those reported for temporal lobectomy. Postoperative deteriorations were broader and stronger in our study than in previous studies. We discuss methodologic differences (sample size, retest interval, extent of resection) and other factors as possible reasons.

Approximately 80% of focal epilepsies originate in the temporal lobes, and anterior temporal lobe resections (TLRs) are the most frequently used surgical treatments in medically resistant focal epilepsies. Apart from unexpected complications, decreased memory functions represent the greatest potential neuropsychological morbidity after anterior temporal lobectomy. Verbal memory functions are particularly at risk in patients undergoing left-sided resection (1–3). Identified risk factors for postoperative memory decline after left anterior temporal lobectomy are older age at time of surgery, later age at onset of seizures, male gender, and better preoperative memory performance (4,5). It has also been shown that patients with normal-appearing hippocampi are at greater risk for memory decline after left anterior temporal lobectomy (6–8). The hippocampal formation, one of the most complex and vulnerable brain structures, is recognized as a crucial brain area subserving human long-term memory. Material-specific relations between the left medial temporal lobe and verbal memory (9,10), and between the right medial temporal lobe and visual/nonverbal memory (11–13) are well described. The relation between verbal memory and the left hippocampus is particularly robust and has been confirmed by different institutions with different approaches (6–8).

Imaging, invasive electrophysiology, clinical data, and pathologic studies indicate the medial temporal structures, in particular the hippocampus, as the primary generator of seizures in temporal lobe epilepsy (TLE) (14). Consequently the question arose, whether a more selective medial temporal resection could result in a comparable seizure control and better cognitive outcome than the standard anterior temporal lobectomy, which would include in patients with pure Ammon's horn sclerosis (AHS) resection of unaffected lateral tissue. In 1984 Spencer modified the standard temporal lobe resection for those patients with medial temporal ictal onset, such that all of the medial structures (amygdala, hippocampus, and parahippocampus) were removed through a limited temporal pole resection. At the same time, Wieser and Yasargil independently devised the more restricted amygdalohippocampectomy through a trans-sylvian approach. In 1988 Wieser reported a favorable seizure outcome in a group of 181 patients after selective amygdalohippocampectomy (SAH): 62% were seizure free, 10% had only rare seizures, and 15% experienced worthwhile improvement (15). Detailed neuropsychological testing was not reported for those patients, but the investigators thought in general that successfully resected patients improved in learning and memory, particularly for material handled by the nonresected side. However, they also pointed out that postoperative memory declines may occur in patients without worthwhile improvement of seizures.

In the following years, several groups published results about the cognitive outcome after SAH. Goldstein and Polkey (16) compared several parameters of verbal and nonverbal memory performance (Benton Visual Retention Test, Rey Osterreith Complex Figure, Digit Span, Corsi Block Tapping Test, WMS-R verbal subtests) in 23 SAH patients and 19 TLR patients 1 to 4 months after surgery. Overall, few scores distinguished between the different surgical procedures for cognitive outcome. Only for right-sided cases did they find some evidence that SAH had a better memory outcome than TLR. Pauli et al. (17) compared the effects of standard TLE, TLRs, and SAH on verbal and nonverbal subtests of the Wechsler Memory Scale–Revised in a group of 69 patients 1 to 2 years after surgery. They reported significant verbal memory declines only after standard resections, but not after SAH or TLRs in the dominant hemisphere. This decline was attributed to the extended neocortical excisions in the standard resections. Helmstaedter et al. (18,19) presented two studies evaluating patients with temporal lobe epilepsies 3 months and 1 year after different left temporal resections. Patients with mesial pathology underwent TLR or SAH (n = 43 and 36), patients with lateral pathology underwent cortical lesionectomy (n = 16 and 11). A subgroup of these patients (n = 15) has been included in the present study. Both TLR and SAH led to a significant deterioration in different aspects of verbal memory assessed with the VLMT, a task similar to the Auditory Verbal and Learning Test of Rey (AVLT). The specific patterns of changes led to the conclusion that left temporolateral and temporomesial resections have different qualitative rather than different quantitative effects on verbal memory (i.e., cortical resections affected aspects of short-term memory, whereas mesial resections affected aspects of long-term memory). Recent neuropsychological data from the Zurich Hospital on 23 left- and 26 right-sided SAH patients showed statistically significant declines in a verbal learning task (AVLT) 3 months after left SAH (20). An increased risk for memory loss was associated with atypical language dominance in this study.

The effects of right SAH on nonverbal memory have been less extensively investigated. Helmstaedter et al. (18,19) included only left resected patients and Regard et al. (20) evaluated only verbal memory. In a previous study, we compared patients with right TLE with and without right hippocampal damage (n = 15 and 13) in their nonverbal memory performance assessed with the DCS-R (13). A subgroup of these patients (n = 12) also is included in the present study. In agreement with other studies (11,12), our results indicated hippocampal damage as the major cause of nonverbal memory deficiencies in preoperative patients with right TLE. However, right SAH did not lead to a further decline, and this is consistent with the findings from Pauli et al. (17), and Goldstein and Polkey (16), who found no significant deteriorations after right SAH in nonverbal memory tests. All in all, the available data indicate that right hippocampal resections do not result in nonverbal memory declines.

On the whole, the published studies reported data from relatively small groups of patients, and the results are equivocal. Some studies reported verbal memory declines after left SAH (16,18,19,20), but others did not (15,17), and only few studies investigated the relation between nonverbal memory and the right mesial temporal lobe. It is not clear to what extent an SAH can lead to a significant decline in memory functions. If SAH leads to memory impairment, it would be of interest to investigate which memory functions in particular are at risk of becoming impaired and which determinants of the memory outcome can be discerned. Our own previous studies did not extensively analyze determinants of memory outcome. In the present study, we investigated the effects of left and right SAH on memory in a large sample of 140 patients. Measures of material-specific learning and memory were assessed before and 3 months after SAH. In addition to group statistics, we present analyses that are based on the individual postoperative changes. Special questions (e.g., concerning the effects of intrahippocampal depth electrodes and complications) are treated in separate analyses.



The subjects were 140 patients with medically refractory epilepsy who had undergone SAH at the University Hospital of Bonn since 1990. Only right-handed patients were included. Preoperative diagnostics comprised in all patients high-resolution magnetic resonance imaging (MRI), surface ictal and interictal electrophysiologic recordings, seizure semiology, and neuropsychological examination. Intracranial recordings were applied in 53 patients (depth electrodes and subdural strips in 13 right and 30 left TLE, subdural strips in one right and two left TLE, and depth electrodes in two right and five left TLE patients). It is important to point out that these patients did not necessarily show hints of a bilateral abnormality. For an intracranial investigation, at least one of the following criteria must be met: (a) inconclusive results of ictal, surface EEG recordings due to artifacts, recording of generalized or multifocal ictal EEG patterns in patients in whom there is a hypothesized monofocal epilepsy, absence of recognizable ictal EEG patterns, or ictal onset incongruent with the findings on MRIs or seizure symptoms; (b) no detectable lesion observed on a standard high-resolution MRI in addition to other findings suggesting a temporal lobe seizure onset on either side; and (c) suspicion of a bilateral origin of the seizure or of a major functional impairment of the contralateral temporal lobe because of the results of imaging, electrophysiologic, and neuropsychological studies.

The excisions were on the left in 66 patients and on the right in 74 patients. The approach was trans-sylvian, as described by Yasargil et al. (21). The hippocampus and the parahippocampal gyrus were usually resected en bloc, and the resection reached at least the middle of the cerebral peduncle at its widest diameter (22). The amygdala was resected as completely as possible. Seizure outcome was favorable with 76% seizure-free patients 3 months after surgery (78% right and 74% left resected). Patients were classified as seizure free only if no seizure or aura had occurred at any time since surgery (Engel outcome class I). One hundred twenty patients had an AHS (67 right, 53 left), five patients had a dual pathology (AHS + another lesion, two right, three left), and 15 patients had only another lesion but no AHS (five right, 10 left) according to the preoperative MRI and the histopathologic findings.

Groups did not differ in clinical characteristics (Table 1). Preoperative global intellectual functioning was estimated preoperatively by a vocabulary test (23), which correlates well with educational levels and is similar to the widely used NART test (24). Intellectual functioning was average in both groups, and there were no significant group differences. Psychomotor speed was assessed with a letter-cancellation test before and after surgery in 133 patients (25). A multivariate analysis of variance (MANOVA) with the preoperative and postoperative data of the cancellation test, side of surgery as between-subjects factor, and test repetition as the within-subjects factor, was performed. There were no significant group differences (Fside of surgery = 1.8, p = 0.18, Fside of surgery x repetition = 0.334, p = 0.56) but both groups improved in psychomotor speed after surgery (Frepetition = 28.5, p < 0.001).

Table 1.  Demographic and clinical data of the patient groups
 Right SAH
(n = 74)
Left SAH
(n = 66)
  1. SAH, selective amygdalo-hippocampectomy; NS, not significant (p > 0.05) in ANOVA or χ2 test. IQ was estimated by a vocabulary test. Psychomotor speed was assessed with a letter-cancellation test. The cells contain the number of patients for gender and seizure status, and group means (standard deviations in parentheses) for the other variables.

Male/female34/4036/30χ2 = 0.1 NS
Age at surgery (yr)31.7 (9.9)32.8 (10.7)F = 0.38 NS
Age at onset of
  epilepsy (yr)
11.2 (8.5)11.1 (9.8)F = 0.01 NS
IQestimated102.4 (16.6)99.0 (10.4)F = 1.6 NS
Psychomotor speed   
 Preoperatively102.4 (13.9)100.1 (11.7) 
 Postoperatively107.4 (12.8)103.6 (14.3) 
Seizure free after
5849χ2 = 0.33 NS

Assessment of memory functions

Patients underwent a neuropsychological examination of verbal and visual memory functions preoperatively and 3 months after surgery. Parallel test versions were used in the postoperative examination.

Visual memory was assessed by the DCS-R (26), a task that has already been proved to be sensitive to right temporal dysfunctions (13,27). The DCS-R requires repeated learning of nine abstract designs in six consecutive trials. The designs are presented consecutively on separate cards at a rate of 2 s per item. After each presentation, the patient has to reconstruct as many items as possible with five wooden sticks of equal length. Patients are instructed to keep the design and the spatial arrangement identical with the target item. The evaluated parameter was the learning capacity measured as the number of correct reproductions in the last trial. Raw values were transformed into standardized scores (mean, 100, SD, 10) using values of a sample of 102 healthy controls (age mean, 29.8, SD, 8.8; sex: 50 male/52 female; nonverbal learning capacity: mean, 7.5, SD, 1.7).

Verbal memory was assessed with the Verbal Learning And Memory Test (28), a German version of the AVLT, which requires learning a word list (15 unrelated concrete words) in five consecutive trials, free recall after learning an interference list, free recall and recognition after a time interval of 30 min. Distracters in the recognition trial were words from the interference list, semantically and phonetically similar words. The parameters of interest were learning capacity (total number of words over all five learning trials), delayed free recall, and recognition performance (correctly recognized words minus false positives). The test results were standardized according to normative data of 200 healthy controls (age: mean, 27.9, SD, 8.7; sex: 158 male, 42 female; verbal learning capacity: mean, 54.8, SD, 8; delayed free recall: mean, 11.5, SD, 2.6; recognition performance: mean, 13.4, SD, 1.9).


Group differences

Memory performance before and 3 months after surgery is reported in Table 2. Inspection of the preoperative results indicates a group difference in verbal memory with impaired performance (standardized scores <85) preferentially in the left group. Group differences are pronounced postoperatively, because left SAH patients score postoperatively far below average (2–3 standard deviations below the mean). Nonverbal memory performance is slightly impaired preoperatively in both groups, with no apparent worsening attributable to surgery.

Table 2.  Memory performance before and 3 months after SAH
 Right SAHLeft SAH
 Preop.3 mo postop.Preop.3 mo postop.
  1. Cells provide the mean standardized scores with the standard deviations in parentheses. Standardized scores have a mean of 100 (SD, 10).

  2. SAH, selective amygdalohippocampectomy.

Verbal memory    
 Learning91.1 (12.6)90.1 (12.2)83.9 (9.5)75.7 (11.7)
 Delayed recall91.4 (13.0)87.3 (13.4)79.3 (10.2)69.1 (10.9)
 Recognition90.7 (19.4)86.4 (18.2)76.1 (23.1)50.8 (28.5)
83.9 (15.2)82.6 (16.0)84.1 (14.7)84.5 (15.8)

A MANOVA was used to examine the pre–post differences statistically. The memory scores (verbal learning, delayed free recall, and recognition, nonverbal learning) were the dependent variables, test repetition was the within-subjects factor, and side of surgery (left/right), gender (male/female), and seizure outcome (seizure free/not seizure free) were the between-subjects factors. This rather complex statistical approach was chosen because it has a far better power than multiple t tests and ANOVAs and because there were considerable positive intercorrelations between the memory measures (Pearson correlations: between the verbal memory parameters, 0.65 < r < 0.77 with all p values <0.001; nonverbal learning capacity with the verbal memory parameters 0.25 < r < 0.4 with all p values <0.01).

The MANOVA yielded a significant main effect of the side of surgery (Hotellings T = 0.47; exact F = 14.7; p < 0.001), which was due to a significant lower performance of the left group in all parameters of verbal memory (with all F values >18.6 and p < 0.001). No effects of the side of surgery were found for visual memory (F = 1.7; p = 0.2). There was a significant effect of test repetition (Hotellings T = 0.47; exact F = 14.6; p<0.001) and significant interactions between test repetition and side of surgery (Hotellings T = 0.25; exact F = 7; p < 0.001) and between test repetition and gender (Hotellings T = 0.09; exact F = 2.8; p < 0.027). Univariate analyses indicated a significant loss in all parameters of verbal memory (all F > 12.4 with p < 0.01). This was indeed due to the left-resected patients, because the interaction of test repetition and side of surgery also was significant for all verbal memory parameters in the univariate analyses (all F values >6.3 with p < 0.014). The interaction of test repetition and gender was significant for the verbal learning capacity (F = 4.7; p = 0.03) and the delayed verbal recall (F = 10.5; p = 0.03); women deteriorated more than men. As Fig. 1 illustrates, this is particularly true for left-sided resections. Left-side resected women scored slightly higher than men preoperatively, but slightly lower than men postoperatively. No significant effects were found for seizure outcome.

Figure 1.

Interaction of test repetition and gender for verbal learning capacity and delayed free recall. In the left selective amygdalohippocampectomy (SAH) group, women score higher than men preoperatively and lower than men after surgery. In the right-SAH group, performance changes were similar for men and women.

Individual changes in memory performance

Decisions about the significance of individual postoperative changes were based on the test–retest data of 85 nonsurgical patients (50 male, 35 female) with complex partial seizures who were on the waiting list for epilepsy surgery and had been tested twice. The mean retest interval in this group was 11.6 months (SD, 6.2); mean age at time of the first test was 33.3 (SD, 13.6); and mean IQest was 102.1 (SD, 14.2). The 90% confidence intervals for the before and after difference scores were calculated following the proposals of Hermann et al. (10), by taking into account the retest reliabilities of the nonsurgical group and correcting for retest effects (estimated by the difference in scores for the nonsurgical group). Individual memory changes in the surgical groups were classified as being unchanged if they were within this confidence interval and worse or improved if they outreached the confidence interval. Table 3 shows that postoperative losses in verbal memory were more frequent after left SAH (24–51% vs. 3–32%). Postoperative losses in nonverbal memory were not different for right or left resections (30 vs. 24%). There were some gains in nonverbal memory irrespective of the side of surgery (14 and 12%).

Table 3.  Individual changes in memory after left or right SAH
  1. SAH, selective amygdalo-hippocampectomy. Cells provide the percentage of patients in the left or right SAH group showing a loss, no change, or a gain in their postoperative performance compared to their preoperative performance.

Verbal memory      
 Delayed recall5132476127
Nonverbal memory243062581412

To evaluate whether these proportions are statistically significant, we calculated binomial tests assuming—based on the 90% confidence interval—that a change in either direction should occur by chance in 5% of the cases. The binomial tests indicated for both groups that the number of losses in all verbal memory parameters, except for verbal recognition in the right SAH group, was significantly higher than expected (p < 0.001). The number of losses in nonverbal memory also was significantly higher than expected for both groups (p < 0.001). The number of gains in nonverbal memory also was significantly higher than expected for both groups (p < 0.05). These gains were not related to the postoperative seizure situation or to changes in psychomotor speed (χ2 < 0.28; F = 0.7 with p > 0.5).

Outcome prediction

Multiple-regression analyses were undertaken with the postoperative values of the memory measures as the dependent variables to identify risk factors for postoperative losses (Forward-selection, inclusion criterion, F > 3.83 with p ≤ 0.05). Preoperative memory scores, side of resection, age at the onset of epilepsy, age at surgery, gender, evaluation with intrahippocampal depth electrodes, and seizure outcome were the independent variables. For the purpose of the calculation, vectors were arbitrarily assigned to side of resection (1 = right, 2 = left), gender (1 = male, 2 = female), evaluation with depth electrodes (1 = yes, 2 = no), and seizure outcome (1 = seizure free, 2 = not seizure free). The results are provided in Table 4. In all parameters of verbal memory, lower postoperative performance was predicted by a lower preoperative performance, left-sided surgery, and an older chronologic age at surgery. The predictors for postoperative verbal memory performance accounted for a high proportion of variance (between 52 and 63%). Postoperative nonverbal memory was predicted only by the preoperative performance accounting for 36% of variance.

Table 4.  Predictors of postoperative memory
Analysis of
Predictorst value
  1. a  p < 0.05; bp < 0.01.

Postop. verbal memory   
 LearningF = 48.8Preop. level7.6b
 R2 = 0.52Side of surgery−5.8b
  Age at surgery−2.4a
 Delayed free recallF = 56.9Preop. level6.9b
 R2 = 0.56Side of surgery−6.0b
  Age at surgery−2.8b
 RecognitionF = 74.5Preop. level7.8b
 R2 = 0.62Side of surgery−7.9b
  Age at surgery−3.3b
Postop. nonverbal memoryF = 77.7Preop. level8.8b
 R2 = 0.36  

Additional analyses


Surgical complications (30) occurred in 16 patients (nine left, seven right). In the right SAH group, patients with complications had, compared with patients without complications, a significantly stronger decline in verbal learning capacity (–10.5 vs. –0.03), delayed free recall (–13.7 vs. –3) and recognition (–18 vs. –3.7) (F > 5.4 with p < 0.024). Table 5 lists the right SAH patients with complications in detail. The extent of memory declines varied considerably, and one patient had no verbal memory declines at all. Of course, a deep vein thrombosis probably did not cause memory declines, but other events could be identified as probable reasons in these patients. For most complications, it is difficult to determine whether left temporal lobe functions were directly affected. It is conceivable that complications could result in a generally slower recovery.

Table 5.  Memory change in right SAH patients with complications
PatientPostop. memory changeDescription of the complication
  1. Memory change is presented as the difference of the pre- and postoperative standardized score; a negative value indicates a loss, rec., recognition; IAT, intracarotid amytal test; SGS, secondarily generalized seizures.

1−15.00−34.62−57.89Deep vein thrombosis, but had an infarction of the left artery cerebri media during the IAT
2−32.50−46.15−42.00Postoperative meningitis
3−1.25−3.850.00Postoperative meningitis
4−13.75−3.85−26.32Deep vein thrombosis, but during the presurgical evaluation (after the preoperative neuropsychological examination), had 12 SGS within 20 h
5−7.500.000.00Postoperative meningitis
6−5.00−11.54−5.26Intracerebral hematoma, left-sided hemiparesis, vigilance problems, and frequent headache after surgery
71.253.855.26Cyst in the resection cave

There was no difference in the left SAH group between patients with complications and patients without complication (F < 0.32; p > 0.5). One must be cautious in interpreting these results because the number of patients with complications is (fortunately) relatively small. If we exclude these patients temporarily, the results of the MANOVA (see section, Group differences) and the binomial tests (see section, Individual changes) are unchanged. Thus complications slightly increased the risk in patients with right SAH for a decline in verbal memory, but they cannot explain the above described results.

Ammon's horn sclerosis

The presence of an AHS has been previously related to preoperative memory deficits, particularly in patients with left temporal lobe seizures (6,7). As a consequence, patients with AHS may have a better outcome than patients without sclerosis, because the latter have more to lose. To evaluate this subject, we further subdivided left and right SAH in a group comprising patients without AHS and another group comprising patients in whom the sole pathology was an AHS (excluding patients with dual pathology). Left-SAH patients without AHS (n = 10) more frequently had a loss in verbal learning capacity (χ2 = 5.9, Fisher's exact test, one-tailed; p = 0.012), and as a trend in verbal delayed free recall (χ2 = 3.1, Fisher's exact test, one-tailed; p = 0.057) than patients with AHS (n = 53). For right-SAH patients, there was no difference in nonverbal learning capacity between patients with AHS (n = 67) and patients without AHS (n = 5; χ2 = 2.7, Fisher's exact test, one-tailed; p = 0.13). Of course, one must be cautious to interpret these results because there is only a small number of patients without AHS!

Intrahippocampal depth electrodes

Intracranial recordings included longitudinal intrahippocampal depth electrodes (IDEs) in 50 patients, and it is of considerable clinical relevance to weigh the use of this diagnostic aid against the potential neuropsychological costs. Evaluation with IDEs was no significant predictor in the regression analyses described earlier. However, particularly memory performance associated with the unaffected contralateral temporal lobe may be at risk for deterioration when bilateral IDEs are inserted. Of special interest is, therefore, a comparison of right-SAH patients with/without IDEs (n = 59 and 15) with respect to verbal memory and a comparison of left-SAH patients with/without IDEs (n = 35 and 31) with respect to nonverbal memory. A multivariate analysis of variance (MANOVA) was performed separately for the left- and right-SAH groups with the verbal memory difference scores (postoperative score – preoperative score) as the dependent variables and IDEs (yes/no) as between-subjects factor. Because we expect larger losses for the patients with IDEs, a directional test with p ≤ 0.1 should be used. On this level of significance, the MANOVA resulted in a significant group difference (Hotellings T = 0.1, F = 2.2, p = 0.09) for right SAH but not for left SAH (F = 0.53, p = 0.66). Univariate post hoc analyses indicated a difference in recognition (with IDEs, –10.9; without IDEs, –3; F = 3.3, p = 0.07), but not in learning capacity (with IDEs, 1; without IDEs, –1.5; F = 0.7, p = 0.4) or delayed free recall (with IDEs, –3.3; without IDEs, –4.4; F = 0.01, p = 0.9). No significant group differences were found for the nonverbal memory difference scores in a univariate analysis of variance (ANOVA: right SAH, F = 0.003; left SAH, F = 0.02; all values of p > 0.8).

Predictors of verbal memory change in the left-SAH group

Given the widespread concern regarding patients with left temporal lobe epilepsy and the high number of patients with a verbal memory decline in this study, additional analyses just on the left-SAH patients seemed to be justified. We computed multiple-regression analyses with the difference scores in verbal memory (preoperative minus postoperative) as dependent variables. Independent variables were the respective preoperative verbal memory score, preoperative nonverbal memory performance, age at the onset of epilepsy, age at surgery, gender, evaluation with IDEs, and seizure outcome. Significant predictors for a loss in verbal learning capacity were a higher preoperative verbal performance, a lower preoperative nonverbal memory score, and a later onset of epilepsy (R2 = 0.34). Significant predictors for a loss in delayed free recall and in recognition performance were a higher preoperative performance and an older age at surgery (R2 = 0.34 and 0.17).


The technique of SAH was originally developed in epilepsy surgery to spare unaffected brain tissue from surgery and thus to minimize the negative cognitive consequences of temporal lobe surgery. The results of previous studies, however, are equivocal in this regard. Some studies reported no memory declines after SAH (15,17), but others did (16,18–20). The present study evaluated memory performance after SAH in a very large sample of patients with mesial TLE. The group statistics clearly indicate that SAH can lead to significant verbal memory declines, particularly in left-resected patients. Analyses of individual changes gave evidence of a loss in verbal memory in ≤51% of the left SAH group and ≤32% of the right-SAH group. General predictors of postoperative memory were similar to those reported for patients who underwent anterior temporal lobectomy (4,5). Poorer postoperative verbal memory was associated with a left-sided surgery, a lower preoperative performance level, and an older age at surgery. A low preoperative performance level often indicates damage of the corresponding structure and is usually associated with a relatively stable pre- to postoperative performance because patients can not lose much more, when the damaged structure is removed. When the postoperative decline was considered, left-SAH patients were especially at risk when they had a higher preoperative performance in the respective score, a later onset of epilepsy, and no AHS. All these aspects indicate a relatively intact preoperative functioning of the hippocampus, and it is evident that such patients are at a greater risk to lose. Declines in verbal memory were associated, in left-SAH patients, with an older age at surgery and a lower preoperative nonverbal memory performance, probably indicating worse capacities for compensation. Patients with right SAH were at a slightly higher risk for a postoperative loss in verbal memory when they had IDEs in the presurgical evaluation or complications. Both aspects could indicate damage to the left temporal lobe or hippocampus, which could be temporary and therefore should be reevalutated in a long-term follow-up. A previous study in our hospital found no evidence of a verbal memory decline in right-SAH patients with IDEs (31).

The results concerning nonverbal memory were less clear. There were no group differences according to the side of surgery. Postoperatively, gains were evident in ∼13% independent on the side of resection, the postoperative seizure status, and the change in psychomotor speed (both groups improved postoperatively). However, there were also declines in nonverbal memory in ∼27% independent of the side of surgery. Postoperative nonverbal memory performance was predicted only by the preoperative score.

Different from previous studies, our results indicated a gender effect in verbal delayed free recall. Left-SAH women were better than men preoperatively and had stronger declines after surgery than did men. The preoperative advantage of women in declarative verbal memory has been described before and corresponds well to the classic fact that women usually perform slightly better in language-production tasks (32,33). The stronger postoperative decline of women, however, contradicts earlier results in patients with lobectomy, which showed that men are at greater risk than women in several series (5,34–36). It could be substantial in this regard whether lateral temporal cortex is resected. However, we are cautious to interpret this result, unless it has been replicated in other studies with SAH patients.

Deteriorations after left SAH concerned all parameters of verbal memory (learning, delayed free recall, and recognition). These results thus indicate a less selective deterioration than that reported in most previous studies, which found postoperative declines after left SAH, especially in long-term aspects, but not in short-term aspects of verbal memory such as immediate recall or learning performance (16,18,19). A comparison of the extent of change with that in other studies is possible in only a few cases. Regard et al. (20) applied the same verbal memory test and presented the results in raw values. A comparison of these results indicates a stronger decline in the present study (verbal learning capacity, –6.6 vs. –4.6; delayed free recall, –2.6 vs. –0.8). A comparison with our own previous studies (18,19) also indicates a stronger decline in the present study in delayed free recall (–2.7 vs. –1.5). Pauli et al. (17) and Goldstein and Polkey (16) applied different tests, and the results were not described in transformed standardized values, which would have facilitated such a comparison. In the study by Wieser (15), the neuropsychological data were not presented. A direct comparison with these latter studies is therefore not possible.

The number of patients with a postoperative memory decline in our study is unexpectedly high and requires an explanation. Seizure outcome (76% seizure free) appears slightly better in our group than in other studies. Wieser (15) reported 62%, and Pauli (17) reported 52% seizure-free patients. Goldstein and Polkey (16) and Regard et al. (20) gave no specification about the percentage seizure free. As regards the surgical approach, the extent of the hippocampal resection might have caused the good seizure outcome in our group, at the expense of greater memory declines. It is known that total hippocampectomy is associated with superior seizure outcomes compared with partial hippocampectomy. However, in other studies, the extent of the hippocampal resection was comparable (2–4 cm). Furthermore, several studies found no relation between memory decline and the extent of a unilateral hippocampal resection (37–39). The parahippocampal gyrus and the amygdala are removed as completely as possible in our medical center. The medial nuclei of the amygdala mostly remain untouched, but the lateral nuclei of the amygdala are always removed completely. This might be an important point because the basolateral nucleus of the amygdala is known to enhance memory consolidation through the release of glucocorticoids (40). However, it is very difficult to compare the exact extent of resections carried out in different centers. Strictly speaking, the extent of the resection must be certified by a postoperative MRI in every patient to be sure. Conversely, because exact comparisons of the resection size are not available, it cannot be excluded as a possible reason for differences in the cognitive postoperative outcome. This applies also to other clinical variables such as the severity of seizures and the degree of SAH, which have been related to memory outcome (41,42). Another factor that recently came to our attention as a possible source of postoperative memory decline is the collateral damage of temporolateral tissue adjacent to the surgical approach via the sylvian fissure (43). In a recent study, Dupont et al. (44) reported after SAH a significant worsening of the hypometabolism on the ipsilateral temporal pole that was not resected! This could be an explanation for the observed decline in aspects of memory that have previously been associated with neocortical temporal structures (16,18). Our respective data are very preliminary and not yet published, but they indicate that this is an issue that should be considered in more detail in the future.

Finally, the postoperative retest interval might also contribute to differences in reports of different centers. Short intervals might still reflect direct postoperative effects that will remit later on. All other studies that reported at least some decline in memory after SAH had comparable postoperative retest intervals of 1–4 months (16,18,20). The only studies that found no decline had a longer retest interval of 1–4 years after surgery (15,17). In our own early series, however, the rather selective deficits persisted 1 year after surgery (19). Finally, the different sample sizes must be discussed. The large sample size in our study probably enabled us to get a better insight into the postoperative dynamics. Random selection effects usually exert stronger influences in smaller samples. Furthermore, large numbers mean a higher power in statistical testing.

In summary, this study revealed less selective and stronger memory declines after SAH than did previous studies. Longer-term follow-ups might reveal some recovery in the identified memory pathology. Future research is necessary that investigates the influence of other possibly relevant variables such as the degree of SAH, severity of seizures, extent of resection, and surgically induced damage to collateral tissue. The gender effect that appeared in this study should also be cross-validated in another study in patients with SAH.