Verbal memory decline from hippocampal depth electrodes in temporal lobe surgery for epilepsy

Transcription

1 FULL-LENGTH ORIGINAL RESEARCH Verbal memory decline from hippocampal depth electrodes in temporal lobe surgery for epilepsy * Hanna Ljung, Arto Nordlund, * Maria Strandberg, #Johan Bengzon, and * Kristina K allen SUMMARY Hanna Ljung is a clinical neuropsychologist and doctoral student at Lund University. Objective: To explore whether patients with refractory mesial temporal lobe epilepsy risk aggravated verbal memory loss from intracranial electroencephalography (EEG) recording with longitudinal hippocampal electrodes in the language-dominant hemisphere. Methods: A long-term neuropsychological follow-up (mean 61.5 months, range months) was performed in 40 patients after ictal registration with left hippocampal depth electrodes (study group, n = 16) or no invasive EEG, only extracranial registration (reference group, n = 24). The groups were equal with respect to education, age at seizure onset, epilepsy duration, and prevalence of pharmacoresistant temporal lobe epilepsy (TLE; 75%) versus seizure freedom (25%). Retrospective neuropsychological data from preoperative surgical workup (T1) and prospective followup neuropsychological data (T2) were compared. A 1 SD intrapatient decline was considered as clinically relevant deterioration of verbal memory. Results: Significant decline in verbal memory was seen in 56% of the patients in the study group compared to 21% in the reference group. At T1, there were no statistical between-group differences in memory performance. At T2, between-group comparison showed significantly greater verbal memory decline for the study group (Claeson Dahl Learning and Retention Test, Verbal Learning: p = 0.05; Rey Auditory Verbal Learning Test, Total Learning: p = 0.04; Claeson Dahl Learning and Retention Test, Verbal Retention: p = 0.04). An odds ratio (OR) of 7.1 (90% confidence interval [CI] ) for verbal memory decline was seen if right temporal lobe resection (R TLR) had been performed between T1 and T2. The difference between groups remained unchanged when patients who had undergone R TLR were excluded from the analysis, with a remaining aggravated significant decline in verbal memory performance for the study group compared to the reference group. Significance: Our results suggest a risk of verbal memory deterioration after the use of depth electrodes along the longitudinal axis of the hippocampus. Until this issue is further investigated, caution regarding depth electrodes in the language-dominant hemisphere hippocampus seems advisable. KEY WORDS: TLE, Verbal memory, Invasive EEG, Hippocampal depth electrodes. Therapy-refractory temporal lobe epilepsy (TLE) is a major concern in epilepsy treatment, and temporal lobe resection (TLR) is the most common form of adult epilepsy surgery. Long-term follow-up shows up to 70% seizure freedom after TLR. 1 However, resection of the temporal lobes is not without risk for patients. Memory deficits, which in Accepted September 30, 2017; Early View publication November 3, *Department of Neurology and Rehabilitation Medicine, Lund University Hospital, Lund, Sweden; Division of Clinical Sciences Helsingborg, Department of Clinical Sciences Lund, Faculty of Medicine, Lund University, Lund, Sweden; Division of Neurology, Department of Clinical Neurosciences Lund, Faculty of Medicine, Lund University, Lund, Sweden; Institute of Neuroscience and Physiology, Sahlgrenska Academy at the University of Gothenburg, M olndal, Sweden; Division of Neurosurgery, Department of Clinical Neurosciences Lund, Faculty of Medicine, Lund University, Lund, Sweden; and #Department of Neurosurgery, Skane University Hospital, Lund, Sweden Address correspondence to Hanna Ljung, Department of Neurology, Lund University Hospital, Lund, Sweden. hanna.ljung@med.lu.se Wiley Periodicals, Inc International League Against Epilepsy 2143

2 2144 H. Ljung et al. Key Points This study showed a risk of verbal memory deterioration from longitudinal hippocampal depth electrodes in the language-dominant hemisphere Verbal memory tests from patients with mesial TLE were compared to presurgical performance, and to the reference group evaluated for surgery without hippocampal depths At long-term retesting, 56% of the patients in the study group and 21% of the reference group had a clinically significant decline in verbal memory This difference supports our hypothesis that depth electrode trajectories can cause persistent disruption of memory circuits in the medial temporal lobe structures many cases are present presurgically because of the epilepsy per se, can be aggravated by surgery. 2 The likelihood of seizure freedom and the risk of postsurgical memory decline are evaluated during the extensive presurgical workup, which includes extracranial electroencephalography (EEG), structural magnetic resonance imaging (MRI), and neuropsychological assessment. Studies on long-term memory show a stable outcome 2 10 years after surgery at the group level. 3 Although some patients have memory deficits, verbal memory deterioration is seen in up to 63% after left TLR (L TLR) versus 25% after right TLR (R TLR). 4 Progressive postsurgical memory deterioration is most common when seizure freedom is not achieved. 5,6 In addition, a decline in verbal naming occurs in up to 50% of all patients after TLR of the language-dominant hemisphere. 4 Invasive EEG monitoring is part of the presurgical workup in 20 30% of all cases. Mesial temporal lobe coverage is achieved by subdural strips or by depth electrodes. 7 For detection of seizure onset in the medial temporal lobes, hippocampal depth electrodes can be implanted unilaterally or bilaterally, by either a parasagittal or an orthogonal approach. 8 Parasagittal placement means along the longitudinal axis of the hippocampus with anterior contacts in the entorhinal cortex, whereas the orthogonal technique lets the depth electrodes enter the temporal lobes laterally. In a major retrospective study from 2012, 260 patients were examined for surgical risks from subdural or depth electrodes. 9 The authors used a modified complication grading system from 0 (no complications) to 4 (persisting neurological deficit >12 months) or 5 (death related to the invasive workup). The study showed hippocampal depth electrodes to be associated with significantly lower surgical overall risks compared to other types of intracranial electrode coverages in the study. Nevertheless, the authors stressed the importance of further studies on the neuropsychological risks associated with the use of intracranial electrodes. To our knowledge, there are only two studies on the safety of invasive EEG with respect to memory function. Both studies examined surgical candidates for memory impairment before and after invasive EEG registrations and found no evidence of memory deterioration caused by depth electrodes placed in the hippocampus. 10,11 Neither of these studies followed the patients for >11 months, and both contained only a few patients who had been examined with hippocampal depth electrodes. The aim of our study was to explore whether longitudinally placing depth electrodes in the hippocampus of the language-dominant hemisphere carries a risk of verbal memory deficits. Methods Patient selection We performed a long-term neuropsychological follow-up of patients with refractory mesial TLE who were eligible for TLR at Skane University Hospital in Sweden. Patients who had undergone workup for TLR between 2003 and 2014, who were over 18 years of age, and who had a level of general cognitive ability >IQ 60 were identified through a retrospective review of medical records (Fig. 1). The study group comprised patients examined with hippocampal depth electrodes in the language-dominant hemisphere, whereas the reference group comprised patients with no invasive electrode placement during their presurgical workup. Patients with ongoing psychosis, progressive neurologic comorbidity, an estimated IQ <60, or which had undergone L TLR were excluded. L TLR patients were excluded because we wanted to examine the possible effects on verbal memory from the depth electrodes per se, not from resective surgery in the dominant temporal lobe in general. Patients prescribed topiramate or zonisamide also met exclusion criteria (n = 2 patients in the reference group), to avoid confounding from drugs with a negative impact on verbal memory performance. Figure 1 shows the selection process. Forty-four patients were eligible for the study group due to invasive video-electroencephalography (veeg) during presurgical workup: 14 had been investigated with subdural strips and 30 with hippocampal depth electrodes. Four of the 30 patients with hippocampal depth electrodes had combined investigations with a grid over the lateral aspect (right n = 3, left n = 1). Thirteen of the 44 patients proceeded to L TLR, four to R TLR, and the remaining 13 had no resective surgery. Seventeen patients with left hippocampal depth electrode implantations were left for inclusion, and all but one provided their informed consent. Finally, the study group comprised 16 patients, of whom 4 had undergone R TLR. Of these 16 patients, 15 had been investigated with bilateral hippocampal depth electrodes and one with a single depth-electrode in the left hippocampus.

3 2145 Hippocampal Depths and Verbal Memory Decline Figure 1. Patient selection process. Study group: TLE patients with hippocampal depth electrodes during invasive investigation for epilepsy surgery. The reference group was balanced against the study group for factors with impact on verbal memory performance. TLR, temporal lobe resection; R TLR, right temporal lobe resection. Epilepsia ILAE A total of 71 patients with extracranial veeg were identified as eligible for the reference group. The selection process aimed at balancing groups according to factors that could possibly affect memory. These factors were ranked based on their putative effect on memory performance (in descending order): R TLR (yes/no); seizure freedom

4 2146 H. Ljung et al. (yes/no); biologic age (10 years vs. respective individual in the study group); age at epilepsy onset (below or above 18 years of age); education; and duration of epilepsy. This procedure left a total of 24 patients in the reference group with either no TLR (n = 18) or R TLR (n = 6) who provided their informed consent for participation in the study. Neuropsychological assessment Neuropsychological assessments were carried out in two sessions, once during the presurgical workup (T1; retrospective data) and once after inclusion in this study (T2; prospective data). The test battery at T1 was nearly identical for all patients despite the long study period (>10 years). Due to changes in presurgical test routines, some tests used at T2 had not been used at T1, thereby limiting the number of tests used in both test sessions. At T2, all patients were tested with an identical test battery. The neuropsychological test battery included the following six tests. First, the Swedish version of the Rey Auditory Verbal Learning Test (RAVLT) consists of a 15-item wordlist test that patients are asked to repeat five times and then recall immediately after each trial, once after a brief distraction of another wordlist, and again after 30 min. 12 Second, in the Claeson- Dahl Learning and Retention Test (CD), patients must recall a 10-word list after a latency period of 15 s after each of the 10 trials or until they correctly recall it twice, and then again after 30 min. 13 Third, Logical Memory, from the Wechsler Memory Scale (WMS), consists of two stories that are to be recalled immediately after each presentation and again after 30 min. 14 Fourth, the Rey Complex Figure Test (RCFT) consists of copying a complex figure that is to be recalled on two occasions. 14 Fifth, in the Brief Visuo-Spatial Memory Test-Revised (BVMT-R), six simple geometric figures must be remembered and recalled on three occasions and then again after 30 min. 14 Sixth and finally, The Warrington Recognition Test for Memory for Faces contains 50 male faces that must be recalled in a multiple choice test. 14 The Claeson-Dahl Learning and Retention Test and the Brief Visuo-Spatial Memory Test-Revised offered parallel versions for retesting, whereas the other memory tests did not. IQ was estimated in all patients, using the Wechsler Adult Intelligence Scale, at the presurgical testing (T1). Aside from measures of memory function, the Boston Naming Test (BNT) was used for testing confrontation naming. The Hospital Anxiety and Depression Scale was used to screen for depression and anxiety. Language lateralization The Edinburgh Inventory of Handedness was used at T2 to confirm hand dominance (Oldfield, 1970). The results from the inventory validated the determination of handedness made during the presurgical workup, regardless of whether patients proceeded to surgery. 15 Patients with declared left-handedness had all been examined for language dominance by either intracarotid amytal test (IAT, or the so-called Wada test ) (n= 4) or functional MRI (n = 5). None of the patients from the R TLR group were right-lateralized in terms of language by either IAT/Wada test or functional MRI, and no patient had unexpected verbal decline or verbal memory deterioration postsurgically. Depth electrode placement A contrast-enhanced CT scan was fused with a preoperative gadolinium contrast enhanced MRI using a stereotactic treatment planning workstation (Leksell SurgiPlan; Elekta AB). A trajectory along the longitudinal axis of the hippocampus was calculated. The target was set at the rostral tip of the hippocampus proper. With the head end of the operating table of the patient elevated approximately 15 degrees and with the neck flexed approximately 30 degrees, a 2.1 mm outer diameter guide cannula was stereotactically directed toward the target point through a 14-mm burr hole over the occipital region. The electrode (eight contacts, 4.4 mm spacing, 1.2 mm outer diameter, PMT Corp.) was placed through the guide cannula, which was then removed. A high-resolution head computed tomography (CT) scanning study the evening after surgery was fused with the patient s preoperative MRI study to control the electrode positions. Figure 2 shows traces of a depth electrode in the left temporal lobe, visualized by a 3T, sagittal, T1-w magnetization prepared rapid gradient echo (MPRAGE) sequence from a control investigation in 2014, 4 years after the operation. Postsurgical MRI Postsurgical MRI was performed in 9 of the 10 patients with R TLR; images showed expected resection of the anterior two thirds of the hippocampal formation in 8 patients. Figure 2. The image shows a visible electrode lesion 4 years after intracranial investigation with longitudinal hippocampal depth electrodes in a 31-year-old woman with refractory temporal lobe epilepsy without mesial sclerosis on MRI in The figure is a 3T, sagittal, T 1 -weighted MPRAGE sequence from a control scan in The trajectory channel from the depth electrode is indicated by arrows. The 1.2 mm electrode was inserted through a burr hole over the occipital region and its pathway goes through the hippocampus. The intracranial investigation proved bilateral independent seizure onsets and this patient was not eligible for resective surgery. Epilepsia ILAE

5 2147 Hippocampal Depths and Verbal Memory Decline In one patient in the reference group, approximately 70% of the hippocampus was unresected. Statistics Descriptive statistics were generated with Microsoft Excel 2016 (Microsoft Corp., Redmond, WA, U.S.A.) and SPSS 22 (IBM Corp., Armonk, NY, U.S.A.). All test data except for the Boston Naming Test and the Rey Complex Figure Test (test parameter Immediate recall) were normally distributed, and statistical comparisons were performed with Student s t-test for independent samples (between-group comparisons) or dependent samples (within-group comparisons). For test data that were not normally distributed, the Mann-Whitney U-test was used. The level of significance was set to Effect size was expressed as Cohens s d. To examine individual clinically relevant differences in memory performance from T1 to T2 we used established definitions for memory deterioration and defined a change of 1 standard deviation (SD) as clinically significant. 16,17 For this we used the same age- and education-related normative reference data that we use in everyday clinical practice. 13,14 To quantify the effect of potential confounding factors, we used a logistic binary regression model. Confounders of interest were those related to individual differences in demographics or to the epilepsy disease per se. Three confounders were identified as possibly relevant. First, biologic age >50 years at T2 (50% of patients in the study group and 17% of patients in the reference group). Second, status post R TLR (25% of patients in both groups), since selected patients are known to experience unexpected verbal memory decline post R TLE. And third, bilateral TLE (50% in the study group and 17% of the patients in the reference group), defined as patients with independent left and right seizure onsets in the temporal lobes verified by presurgical ictal EEG recordings (with a requirement of at least 20% versus 80% seizure onset in either temporal lobe). The latter group is known to be a category with more severe epilepsy. Each confounder was analyzed separately, based on whether individual memory deterioration between test sessions 1 SD was present. Ethics This study was approved by the regional ethical review board in Lund, Sweden. Results Demographics Demographic and clinical characteristics of the study subjects are shown in Table 1. The only significant difference was a higher mean age in the study group. Groups did not differ significantly with regard to seizure control, although the number of patients with generalized tonic clonic seizures was higher in the reference group. There was no difference between groups in the proportion of patients with persistent pharmacoresistant TLE (75%) at long-term follow-up (T2). In both groups, 25% of patients underwent R TLR, leaving 75% ineligible for resective surgery after presurgical evaluation. All patients in the study group with status post R TLR were seizure free at T2, whereas all patients who did not have resective surgery continued to have refractory seizures. In the reference group, one of six patients with status post R TLR continued to have seizures (Engel class III), whereas five of six were seizure free at T2 (Engel class I = 83%). Only one of the 18 patients who did not proceed to resective surgery was seizure-free at T2. Antiepileptic drugs (AEDs) The average numbers of AEDs for each group at T1 and T2 are presented in Table 1. Six percent in the study group and 17% in the reference group had no AED treatment at follow-up. Changes in number of AEDs were the following: (1) in the study group, 9 of 16 patients had an unchanged drug regimen, 6 of 16 had 1 or 2 fewer AEDs at T2 versus T1, and one of 16 had one additional AED at follow-up; (2) in the reference group, 16 of 24 had an unchanged drug regimen; 5 of 24 had 1 or 2 fewer AEDs at T2 versus T1; and 3 of 24 had one additional AED at T2. Memory performance At T1, that is, during the presurgical workup before placement of depth electrodes, there were no significant differences between the groups in any of the memory variables tested (Table 2). At T2, the study group performed significantly worse than the reference group in 4 of 12 memory variables (CD verbal learning, p = 0.05; CD retention, p = 0.04; RAVLT total learning, p = 0.04; RCFT delayed recall, p = 0.03), as presented in Table 3. Effect sizes varied between 0.29 and The clinical relevance of the observed differences is supported by strong effect sizes (>0.60) even in some of the statistically nonsignificant comparisons. Memory performance deteriorated in five of six test variables in the study group, versus two variables in the reference group. However, these differences did not reach statistical significance (Table 3). A post hoc power analysis revealed that the between-group comparisons were underpowered (power per test variable ranging from 0.19 to 0.44). Nevertheless, as shown in Figure 3, deterioration becomes evident when looking at within-group changes in memory performance from T1 to T2. The study group deteriorated by 20% in verbal learning and 21% in verbal retention on the Claeson-Dahl test, whereas the reference group showed only 4% and 1% deterioration, respectively. Comparison was not possible for the Rey Auditory Verbal Learning Test, Logical Memory from the Wechsler Memory Scale, and the Brief Visuo-Spatial Memory Test-Revised, since these tests were not used in clinical routines before 2014.

6 2148 H. Ljung et al. Table 1. Description of demographic data, seizure types and frequencies, time to neuropsychological assessment follow-up, and AED treatment in the study group and the reference group Data category Study group (n = 16) Reference group (n = 24) p-value Sex, n (%) Male 9 (56) 10 (42) Not tested Female 7 (44) 14 (56) Handedness, n (%) Right 11 (69) 20 (83) Not tested Left 5 (31) 4 (17) Education at T2, years Mean SD Median (range) 13.5 (8 20) 12 (8 17) Age at epilepsy onset, years Mean SD Median (range) 18.5 (6 49) 15.5 (1 60) Duration of epilepsy at T2, years Mean SD Median (range) 25.5 (8 51) 17 (6 58) Age at T2, years Mean SD Median (range) 51.5 (32 63) 36.5 (25 61) Seizure types Focal seizure, aware 1 (6%) 1 (4%) Focal seizure, impaired awareness 11 (69%) 16 (67%) GTC a seizures: 2 (13%) 8 (29%) <1/year 1 (6%) 4 (17%) 1/year 1 (6%) 3 (13%) Seizure frequency, n (%) Seizure free 4 (25%) 7 (29%) 1 seizure/week 6 (38%) 11 (46%) 1 seizure/month 5 (31%) 1 (4%) 1 seizure/6 months 1 (6%) 5 (21%) Time from T1 to T2, months Mean SD Median (range) 78 (22 107) 50 (17 111) Number of AEDs at T1 Mean SD Median (range) 2 (1 4) 2 (1 3) Number of AEDs at T2 Mean SD Median (range) 2 (0 3) 2 (0 3) Individual number of AEDs at T2, n (%) No AEDs 1 (6%) 4 (17%) Monotherapy 5 (31%) 1 (4%) Polytherapy (2 or 3 AEDs) 10 (63%) 19 (79%) AED distribution Valproate 5 (31%) 8 (33%) Carbamazepine 7 (44%) 7 (29%) Levetiracetam 8 (50%) 16 (67%) Lamotrigine 8 (50%) 9 (38%) Others b 2 (13%) 7 (29%) GTC, generalized tonic clonic. a Focal seizures evolving to generalized tonic clonic seizures. b Phenytoin, lacosamide, perampanel, oxcarbazepine, gabapentin, pregabalin. The most important outcome parameter in our study was clinical relevance of memory decline between T1 and T2, defined as a change in memory performance 1 SD. Nine (56%) of 16 patients in the study group had a clinically significant deterioration in verbal memory performance at T2 compared to T1, whereas only 5 (21%) of 24 in the reference group showed a corresponding decline. Figure 4 illustrates the logistic binary regression analysis of all patients, that is, the study patients and reference patients as one group. There was a sevenfold risk of verbal

7 2149 Hippocampal Depths and Verbal Memory Decline Table 2. Neuropsychological test results at presurgical testing (T1) Test Study group Reference group p-value CD Verbal Learning CD Verbal Retention RCFT Immediate Recall RCFT Delayed Recall Warrington RMT Boston Naming Test a CD, Claeson-Dahl Learning and Retention Test; RAVLT, Rey Auditory Verbal Learning Test; WMS, Wechsler Logical Memory; RCFT, Rey Complex Figure Test; BVMT-R, Brief Visual-Spatial Memory Test Revised; Warrington, Warrington Recognition Test for Memory for Faces; BNT, Boston Naming Test. Neuropsychological test results at pre-surgical testing (T1) for both groups. Data are given as mean (SD). Note that the Rey Auditory Verbal Learning Test, Logical Memory from the Wechsler Memory Scale, and the Brief Visuo-Spatial Memory Test-Revised were excluded due to missing values; those tests were not used before 2014 due to changes in presurgical test routines. In addition,, note that a higher score on the Claeson-Dahl Learning and Retention Test means a reduced performance. a p memory deterioration after R TLR in our population. Neither biological age >50 nor bilateral TLE was found to increase the risk of verbal memory deterioration in our study population. Conversely, the group with bilateral TLE showed less deterioration in memory performance between T1 and T2. Because R TLR was a strong risk factor for verbal memory deterioration, we also checked for verbal memory performance at T2 in the group of patients who did not proceed to resective surgery (study group, n = 12; reference group, n = 18). Differences between groups in memory performance remained unchanged in almost all aspects (statistically and clinically relevant ones), with a marked decline in memory performance in the study group. Furthermore, we looked at memory performance at T1 (i.e., during presurgical workup) in R TLR patients who were eligible for the reference group during the patient selection process, but excluded for one of the following reasons: no match, declined participation, or could not be contacted (shown in Fig. 1). We found no significant differences in verbal memory performance between excluded patients and the reference group, and we did not find any differences between excluded and study-group patients. Verbal function Figure 3 shows the relative decline in verbal naming capacity in the study group compared to the reference group between T1 and T2. At T1, the study group performed significantly better than the reference group on the Boston Naming Test. At T2, the study group performed equal to the reference group, indicating a significant deterioration (Table 3). Psychiatric status There were no significant differences in either anxiety- or depression-related symptoms between the groups, at either Table 3. Comparison of neuropsychological test results at follow-up (T2) Mean difference SD p-value CI (95%) Effect size (Cohen s d) Study group vs. Reference group CD Verbal Learning a to ( 0.018) 0.67 CD Verbal Retention a to RAVLT Total Learning a to RAVLT Immediate Retention to RAVLT Delayed Retention to WMS Logical Memory Learning to WMS, Logical Memory Recall % to RCFT Immediate Recall a to RCFT Delayed Recall to BVMT-R Total Learning to BVMT-R Recall to Warrington RMT to Boston Naming Test to Change from T1, Study group vs. Reference group CD Verbal Learning to CD Verbal Retention to RCFT Immediate Recall to RCFT Delayed Recall to Warrington RMT to Boston Naming Test a to CD, Claeson-Dahl Learning and Retention Test; RAVLT, Rey Auditory Verbal Learning Test; WMS, Wechsler Logical Memory; RCFT, Rey Complex Figure Test; BVMT-R, Brief Visual-Spatial Memory Test Revised; Warrington, Warrington Recognition Test for Memory for Faces; BNT, Boston Naming Test. Rey Auditory Verbal Learning Test, Logical Memory from the Wechsler Memory Scale, and the Brief Visuo-Spatial Memory Test-Revised were excluded from analysis of the within-group changes; these tests were not used before 2014 due to changes in presurgical test routines. Note also that a higher score on the Claeson-Dahl Learning and Retention Test means a reduced performance. a p 0.05.

8 2150 H. Ljung et al. Figure 3. Intragroup change (%) in neuropsychological performance from presurgical workup (T1) to long-term follow-up (T2). CD, Claeson-Dahl Learning and Retention Test; RCFT, Rey Complex Figure Test; Warrington RMT, Warrington Recognition Test for Memory for Faces; BNT, Boston Naming Test. Epilepsia ILAE Figure 4. Univariate analyses of demographic and epilepsy-related factors. Forest plot of age at T2 ( 50 years old versus >50 years old), R TLR between T1 and T2 (R TLR versus no R TLR), and presence of bilateral TLE verified by ictal EEG (bilateral TLE versus unilateral TLE). TLE, temporal lobe epilepsy; R TLR, right temporal lobe resection. Epilepsia ILAE T1 or T2, based on results from the Hospital Anxiety and Depression Scale. Discussion This study is the first to suggest an increased risk of verbal memory deterioration from longitudinal hippocampal depth electrodes in the dominant hemisphere. Presurgical verbal memory test results from patients with mesial TLE were compared with an otherwise equivalent reference group. At retesting, 56% of the patients in the study group versus 21% of the patients in the reference group had clinically significant memory deterioration. In addition, at follow-up, statistically significant differences between groups were found in 4 of 12 memory variables, differences that correlated with strong effect sizes. Altogether, our results support the hypothesis that depth electrode trajectories can cause persistent disruption of memory circuits in the medial temporal lobe structures. Two previous studies have compared memory performances after R TLR, in patients with or without depth electrodes at presurgical monitoring. 10,11 These studies refuted verbal memory deficits caused by longitudinal implantation in the speech-dominant hemisphere after postoperative testing at 3 and 11 months. Our study differs from these studies by applying long-term neuropsychological follow-up and primarily by examining patients who did not proceed to resective surgery. Our findings also differ in that a decline

9 2151 Hippocampal Depths and Verbal Memory Decline was detected for verbal learning, verbal recall, and naming capacity which has never been demonstrated before. Because the length of time between test sessions exceeded 4 years for both groups, we argue that there was a reasonable time for the effect of test practice to disappear. Time to follow-up was different between groups (p = 0.052), with the study group having a longer followup period (69 months vs. 54 months). This raises the question of whether the deterioration in memory performance in the study group could be partly explained by a longer period of refractory TLE. We question the significance of this issue, since previous studies have shown contradictory results A study by Hermann et al. showed a decline in delayed verbal recall (compared to controls) over a period of 4 years in a subgroup of patients with TLE with several known risk factors for memory deterioration. 18 On the other hand, Helmstaedter and Elger discussed developmental hindrance as an alternative explanation for poor memory performance in people with TLE, making age at epilepsy onset the most important factor for long-term memory performance. 19 The authors concluded that the memory deterioration slope in patients with TLE follows the same course as seen in healthy individuals, although it starts at a lower level in patients with early onset epilepsy. This reasoning was confirmed by Mameniskien_e et al. who found no aggravated deterioration in cognitive performance in people with epilepsy compared to age-matched controls during a 13-year memory follow-up. 20 Multiple factors may affect memory in TLE. For that reason, we identified factors that could have influenced our test results, based on knowledge of the causes of verbal memory decline in epilepsy. The study group and reference group were equal in terms of the following parameters: seizure control, age at epilepsy onset, epilepsy duration, number of AEDs, and level of education, which are known risk factors for memory decline in epilepsy. 3,6,21 Groups were not equal for biologic age, which raised the question of whether the difference between groups in memory performance at T2 could be explained by the age difference. Episodic memory curves decline with age, although not all aspects of this phenomenon are fully understood. 22 Because the effect of biologic age warranted further analysis, a binary regression analysis was carried out to evaluate covariates for memory deterioration in our cohort (Fig. 4). Even though relative to the reference group, a large proportion of patients in the study group were older than 50, results from the regression analyses did not support the suggestion that biologic age >50 years contributed to a clinically significant increased risk of memory deterioration between test sessions. Furthermore, bitemporal epilepsy did not constitute an increased risk for verbal memory decline: These patients deteriorated less in memory performance than patients with unilateral TLE. The last bivariate we examined was the possible impact of R TLR on verbal memory performance. This is an important issue, since there is evidence supporting deterioration in memory performance after right as well as left TLR. 23 Our results confirm the findings of others 21 by showing a sevenfold risk of verbal memory decline after R TLR. However, the proportion of R TLR patients was the same in both groups. Furthermore, between-group differences in memory performance remained unchanged after deselecting the R TLR patients from both groups. In addition, we found between-group differences in visuospatial memory at T2 that were not present at T1. This is a noteworthy finding, although it was not the focus of this study, and it may have different explanations. First, visuospatial memory may have been affected by the depth electrodes that were placed in the nondominant temporal lobe, since 15 of 16 patients in the study group had bilateral depth electrodes. Another possible explanation is that the proportions of patients with R TLE or bilateral TLE differed between groups. The small number of patients is a limitation in this study. Although the study group showed numerically large changes in memory performance between T1 and T2, at the group level and in intrapatient comparisons, these changes did not reach statistical significance. It cannot be ruled out that some of the nonsignificant differences were false negatives, as a post-hoc power analysis revealed that most comparisons were underpowered, most likely explained by the small group sizes. This is because Lund has been the only Swedish surgical epilepsy program to apply longitudinal placement of hippocampal depth electrodes. This study is the first to examine the possible long-term effects of longitudinally placed depth electrodes on memory performance. Our results suggest that this invasive method is potentially harmful, since a doubled risk for clinically significant verbal memory decline was seen in the study group compared to the reference group. This finding must be weighed against other studies on the benefits and potential risks associated with hippocampal depth electrodes. We conclude that larger studies are urgently needed. Until then, we recommend alertness regarding the possible risks when using these electrodes. Acknowledgments The authors thank Hakan L ovkvist for statistical assistance. We also thank the epilepsy surgery team at Skane University Hospital in Lund, Sweden, for their contribution to the research, with special thanks to epilepsy nurse Anna Hugoson and biomedical technician Elisabeth Svensson for assistance with data collection. We thank Isabella Bj orkman-burtscher for assisting us with images of a patient from the study with depth electrodes. Funding This study was supported by independent research grants from the Stig & Ragna Gorthons Foundation, Thelma Zoegas Foundation, funding from Region Skane (Sweden), and the Skane University Hospital.

10 2152 H. Ljung et al. Disclosure None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal s position on issues involved in ethical publication and that this report is consistent with these guidelines. References 1. Edelvik A, Rydenhag B, Olsson I, et al. Long-term outcomes of epilepsy surgery in Sweden: a national prospective and longitudinal study. Neurology 2013;81: Baxendale S, Thompson PJ, Duncan JS. Neuropsychological function in patients who have had epilepsy surgery: a long-term follow-up. Epilepsy Behav 2012;23: Andersson-Roswall L, Engman E, Samuelsson H, et al. Cognitive outcome 10 years after temporal lobe epilepsy surgery: a prospective controlled study. Neurology 2010;74: Dulay MF, Busch RM. Prediction of neuropsychological outcome after resection of temporal and extratemporal seizure foci. Neurosurg Focus 2012;32:E4. 5. Helmstaedter C, Kurthen M, Lux S, et al. Chronic epilepsy and cognition: a longitudinal study in temporal lobe epilepsy. Ann Neurol 2003; 54: Thompson PJ, Duncan JS. Cognitive decline in severe intractable epilepsy. Epilepsia 2005;46: Hedegard E, Bjellvi J, Edelvik A, et al. Complications to invasive epilepsy surgery workup with subdural and depth electrodes: a prospective population-based observational study. J Neurol Neurosurg Psychiatry 2014;85: Mormann F, Osterhage H, Andrzejak RG, et al. Independent delta/ theta rhythms in the human hippocampus and entorhinal cortex. Front Hum Neurosci 2008;2:3. 9. Wellmer J, von der Groeben F, Klarmann U, et al. Risks and benefits of invasive epilepsy surgery workup with implanted subdural and depth electrodes. Epilepsia 2012;53: Fernandez G, Hufnagel A, Van Roost D, et al. Safety of intrahippocampal depth electrodes for presurgical evaluation of patients with intractable epilepsy. Epilepsia 1997;38: Carvajal-Castrillon J, Zapata J, Galeano-Toro LM, et al. [Changes in verbal and nonverbal memory associated with bilateral hippocampal electrode implantation for epilepsy surgery]. Rev Neurol 2012;55: Lezak M, Howieson D, Bigler E, et al. Memory II: batteries, paired memory tests, and questionnaires neuropsychological assessment. New York: Oxford University Press; 2012: Claeson L, Esbj ornsson E, T annerus B, et al. Beskrivning av testet. Claeson-Dahls test f or innl arning och minne. Reviderad version, Psykologif orlaget AB; 1998: Strauss E, Sherman E, Spreen O. Memory a compendium of neuropsychological tests: administration, norms, and commentary. New York: Oxford University Press; 2006: Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 1971;9: Hermann BP, Seidenberg M, Haltiner A, et al. Relationship of age at onset, chronologic age, and adequacy of preoperative performance to verbal memory change after anterior temporal lobectomy. Epilepsia 1995;36: Kneebone AC, Lee GP, Wade LT, et al. Rey complex figure: figural and spatial memory before and after temporal lobectomy for intractable epilepsy. J Int Neuropsychol Soc 2007;13: Hermann BP, Seidenberg M, Dow C. Cognitive prognosis in chronic temporal lobe epilepsy. Ann Neurol 2006;60: Helmstaedter C, Elger CE. Chronic temporal lobe epilepsy: a neurodevelopmental or progressively dementing disease? Brain 2009;132: Mameniskien_e R, Rimsien_e J, Puronait_e R. Cognitive changes in people with temporal lobe epilepsyover a 13-year period. Epilepsy Behav 2016;63: Thompson PJ, Baxendale SA, McEvoy AW, et al. Cognitive outcomes of temporal lobe epilepsy surgery in older patients. Seizure 2015;29: Park DC, Lautenschlager G, Hedden T, et al. Models of visuospatial and verbal memory across the adult life span. Psychol Aging 2002; 17: Gleissner U, Sassen R, Lendt M, et al. Pre- and postoperative verbal memory in pediatric patients with temporal lobe epilepsy. Epilepsy Res 2002;51: