If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Surgery negatively affects memory in the short rather than long term follow-up.
In the long run, memory can recover to baseline levels when seizures are controlled.
Drug load reduction positively affects executive functions.
Age and ageing are recovery limiting factors.
Evidence does not indicate accelerated cognitive decline in the years after surgery.
Epilepsy surgery can be a successful treatment option for temporal lobe epilepsy but there are concerns about accelerated memory decline in long-term follow-up.
161 adult operated (77 right, 84 left temporal resections) versus a heterogeneous group of 208 non-operated patients with focal epilepsies were consecutively recruited and re-evaluated focusing on memory, executive functions, and vocational outcome after follow-up intervals of >5 years (5–22 years, mean 8 ± 3 years).
Major losses in the operated group manifest early, at one-year follow-up. Few patients declined further. Long-term changes after surgery did not differ from those observed without surgery. The factor “surgery” caused verbal memory decline, “seizure freedom” (operated 48%, non-operated 17%) was associated with recovery of verbal memory, and “drug reduction” positively affected the course of executive functions. In terms of the critical function of verbal memory, only 3–17% of seizure-free patients showed long-term decline whereas 16–20% showed improvement (operated and non-operated). Persistent seizures were associated with decline in 12–37% and improvement in 4–12% of patients. Improvement was related to longer retest intervals and was stronger in younger patients. Independent of surgery or seizure outcome, the vocational status remained unchanged or change was negative (22%) rather than positive (3%).
Patients’ cognitive course 5–22 years after surgery is stable and may even be positive if epilepsy is controlled and drug load reduced. Depending on seizure outcome, recovery is more frequently observed than continuing decline. Recovery, however, takes time and age is a limiting factor.
It is generally accepted that epilepsy surgery is an excellent treatment option for achieving permanent seizure control in patients with temporal lobe epilepsies but it bears the risk of new and additional memory impairments after surgery [
Human declarative memory depends on the bilaterally disposed temporomesial structures and the hippocampi in particular. After unilateral temporal lobe surgery, the patient is left with only the contralateral temporal lobe, and, having amnestic patients with bilateral temporal and hippocampal damage in mind, it stands to reason that in the long run patients who have undergone temporal lobe resections are at risk of further and accelerated memory decline through the combination of continuing epilepsy and normal or even pathological ageing.
In 2002, we suggested that accelerated memory decline may be seen after temporal lobe surgery due the interaction of surgical defects with ageing, especially in older patients. This was demonstrated by comparing age regressions of memory before and after surgery [
]. Rebecca Rausch’s group, in 2003, was the first to raise the concern of continuing memory decline in 44 temporal lobe resected patients who had been retested more than nine years after epilepsy surgery [
Longitudinal studies from Germany, Great Britain, the Netherlands and more recently Sweden and Italy, however, do not provide confirmatory evidence of a continuing and accelerated memory decline after surgery. These studies cover retest-intervals between two and ten years and indicate that the course of memory performance after surgery tends to remain quite stable [
]. Another study from this group recently identified older patients with tau pathology who are at risk of postoperative deterioration. This finding indicates that pathological ageing could be involved in accelerated cognitive decline [
Against this background, the present study evaluated the cognitive outcomes of patients who had undergone temporal lobe surgery and who had follow-up evaluations more than 5 years postoperatively. A mixed group of non-operated epilepsy patients with comparable test-retest intervals served as a control group. Vocational outcome at baseline and follow-up served as measures of functional outcome.
Memory performance - which, in past studies, was demonstrated to be sensitive to temporal lobe pathology and surgery - was the major outcome parameter. Extratemporal lobe executive functions were assessed as a control condition for memory.
The aim of the study was to evaluate the impact of five factors on the long-term outcome of memory and executive functions in epilepsy patients:
1. “Seizure control” i.e. controlled epilepsy in seizure free vs. uncontrolled epilepsy in patients who were not seizure free. The prediction would be that better outcomes are achieved with greater seizure control.
2. “Surgery” i.e. resection of temporal lobe brain tissue vs. no surgery. The prediction would be that, at least soon after surgery, there should be greater cognitive decline with surgery than conservative treatment. Memory should be affected in particular.
3. & 4. “Age(ing)” i.e. chronological age and the retest-interval. The prediction would be that greater losses are observed in older subjects, and that cognition declines with longer time intervals. This should concern both memory and executive functions.
5. “Drug change” in terms of reduction of drug load (number of drugs) vs. unchanged or increased drug load. Here the prediction would be that cognition improves / deteriorates with a decrease / increase of the postoperative drug load. Executive functions should be affected in particular.
Drug change was included because there are now two studies in children and two studies in adults which demonstrate positive effects of postoperative drug withdrawal or drug load reduction on cognitive recovery after surgery [
This is a retrospective, longitudinal, observational study. Patients had undergone long term follow-up testing in a natural in/outpatient setting. Long term follow-up in our clinic is not part of clinical routine. The patients described in this study had returned to our centre for different reasons, and testing was not initiated for the purpose of this study but requested by the treating physician for a range of clinical reasons including the assessment of treatment-related cognitive changes or to obtain objective long-term follow-up.
Patients were selected from a clinical database capturing all neuropsychological assessments of epilepsy patients in the department since 1986. The selection criteria comprised the availability of a baseline neuropsychological evaluation (T1) of memory functions in the past and a follow-up visit with neuropsychological testing of memory after an interval of more than 5 years since baseline assessment in non-operated and since surgery in operated patients (T3). Intervals between baseline presurgical evaluation and surgery were mostly short (1–2 months). The majority of the surgical patients had an additional routine follow-up visit with neuropsychological assessment one year after surgery (T2). Patient data were collected over the past 15 years.
Seizure freedom was defined as being completely seizure free for any kind of seizure for more than 1 year prior to assessment.
Drug load was defined as the number of antiepileptic drugs (AED). We have shown previously that this number correlates highly with the defined daily dose (DDD) and equally well with executive functioning as measured by EpiTrack, a screening tool for the assessment of cognitive drug side effects. [
Since there are no normative data for retesting after such long intervals, and in order to have any control for the long-term effects of surgery, a heterogeneous control group of non-operated, medically treated patients (focal epilepsies) with comparably long retest-intervals were extracted from the database.
2.2 Study design and statistics
The study design was longitudinal, retrospective and observational. The aim of the study was to evaluate the differential impact of temporal lobe surgery (yes/no), seizure control (yes/no), drug change (change of the number of AED [
]), age and retest interval on long-term cognitive outcome in terms of memory and executive functions.
Patients’ vocational status was another outcome parameter.
It is important to note that the patient selection criteria and the long collection period could introduce a selection bias. Therefore seizure outcome in operated and non-operated patients was of interest mainly as an experimental condition.
The data were analysed on group and individual level. Group level statistics included separate MANCOVAs of long term change (T1 – T3) in verbal memory, figural memory, or executive function (=intra-subject variables) as a function of “treatment” (operated vs. non-operated), and “seizure control” (seizure free vs. not seizure free at T3,) as inter-subject variables, and, as covariates, “chronological age”, “retest interval”, and “drug change” (change of the number of antiepileptic drugs) were performed.
Since this was essential for the question of whether surgery accelerates memory decline after the standard one year follow-up, the same analyses were calculated comparing the course of these functions between one year postoperative follow-up (T2) and long term assessment (T3) in operated versus T1 and T3 in non-operated patients.
Individual level analyses comprised frequencies of impairments according to normative data, as well as change of performance. For this analysis the surgical group was divided into those operated on the left and right. For operated patients the frequencies of individual changes from baseline (T1) to one year postoperative (T2), from one year postoperative (T2) to long term (T3), and from baseline (T1) to long term (T3) were calculated. For the non-operated patients frequencies of changes between T1 and T3 were calculated. Individuals were categorised as impaired and experiencing cognitive change when their performance was at least one standard deviation below the mean performance of healthy subjects or when their performance in test 1 and test 2 had changed positively or negatively by at least one standard deviation.
Predictors of the vocational outcome were analysed with multiple stepwise regression analysis.
All statistics were calculated with the IBM SPSS Statistic 22 package. Test scores were transformed into age corrected standardized scores (M = 100, SD = 10) according to normative data of 488 healthy subjects (254 female, 234 male, age range 18–93 yrs.) who had undergone memory and executive testing. In the older (60+) subjects included in the normative sample Mild Cognitive Impairment or dementia were excluded by including only subjects with normal Mini Mental State Examination scores.
2.3 Neuropsychological assessment
Verbal and nonverbal capacities in episodic memory were assessed using the VLMT (Verbaler Lern- und Merkfähigkeitstest, a German adaptation of the Rey Verbal Learning Test [
]. The VLMT includes word list learning (15 words) with immediate free recall in 5 learning trials, free recall after distraction, delayed free recall, and recognition; the DCS-R requires design learning (9 designs) followed by immediate recall in 5 learning trials and delayed recognition.
Both tests have been successfully used in several studies of cognitive effects of epilepsy surgery and of temporal lobe surgery on memory in particular and can be considered as predominantly reflecting temporal lobe memory function (for overview see [
In order to condense information yielded by the many memory parameters, two test parameters were selected from each test, which are most representative for total test performance (verbal: learning over 5 trials and remembered words in delayed free recall, figural: learning over 5 trials and supra-span performance in the last learning trial). For each test the two test parameters were transformed into standard values (M = 100, SD = 10) according to normative data from the same 488 healthy subjects, and merged ((score 1 + 2)/2) into summary scores for verbal and figural memory.
2.3.2 Executive functions
From all executive functions tested in the evaluated patients, consistent data sets were available for phonemic verbal fluency (finding words with specific initial letters) and for response inhibition (inverse open reading of rows of 1 and 2, i.e. reading “2″ instead of “1″ and vice versa). Both tests are subtests of the EpiTrack, a tool for the assessment of executive functions and working memory, which is sensitive to frontal lobe dysfunction and negative antiepileptic drug effects in particular [
]. To condense the information, the two scores were standardized using test norms from 488 healthy subjects, summed up and divided by two, resulting in a single standardized summary score for executive functions.
Since executive functions and memory have been standardized in relation to the same group of healthy subjects, the tests could be considered as co-normalized.
As a rule, and since the tests used were deficit-oriented, performance values one standard deviation below the mean (10 points) were considered as impaired. As for individual performance changes, a change of more than one standard deviation was considered as significant when the test scores of T1, T2 and T3 were subtracted from each other.
To reduce practice effects, alternate versions of memory and word fluency tests were used for repeat tests.
161 patients who had undergone epilepsy surgery for temporal lobe epilepsy (77 right, 84 left temporal) fulfilled the inclusion criteria of preoperative baseline assessment, and follow-up memory testing after a time interval of more than 5 years after surgery. The control group consisted of 208 non-operated, purely medically treated patients for whom baseline and follow-up neuropsychological test data were available (for patient characteristics see Table 1).
Table 1Demographic and clinical data of the study groups.
Right TLE surgery
Left TLE surgery
Age at baseline
Age at onset
Duration of epilepsy at T1
Follow up interval
Education ≥ 10 yrs.
Drugs at T1
Drugs at T3
off drug monotherapy polytherapy
3% 20% 77%
13% 40% 47%
12% 39% 49%
14% 41% 45%
Change in number of AEDs T1-T3
Seizure free at T3
Seizures T2 & T3
Employed or in education at T1
Employed or in education at T3
N = number, M /SD = mean/standard deviation.
ANOVA Chi-Square: * = p < 0.05; **= p < 0.01, *** = p < 0.001.
AED = antiepileptic drugs.
T1 T2 T3 = baseline, one year postoperative follow up, long term follow up.
The mean age at baseline was 29 ± 12 years in the operated and 31 ± 13 years in the non-operated group, the oldest age of patients at T1 was 64 and 65 years respectively. The mean age at T3 was 37 ± 12 in the operated and 39 ± 14 in the non-operated group, the oldest patients were 68 and 77 years old respectively at the time of the last follow test. The age at the onset of epilepsy was earlier in the operated group (F = 6.5, p < 0.05), but the groups did not differ significantly in terms of duration of epilepsy.
The average follow-up intervals were 8 ± 3 years in both operated and non-operated patients, and the maximum follow-up intervals were 22 and 20 years respectively.
At baseline (T1) all patients were on AEDs, with close to 68% being on polytherapy in both groups. However, the groups differed significantly at T3. The non-operated group was less likely off AEDs (3% vs. 13%) and more likely to be taking polytherapy (77% vs 47%). As for the change of the drugload over time between T1 and T3, the number of AED was reduced in 31% of all patients, unchanged in 32%, and increased in 37%. Changes were significantly different for operated versus non-operated patients (see Table 1), i.e. increase was more likely in non-operated treated, withdrawal more likely in operated patients (chi² = 23.7, df2, p < 0.001). Complete withdrawal was the exception in non-operated patients (3%) and very rare even after surgery (13%).
At T3 17% non-operated versus 41% operated patients were categorised as seizure free, i.e. they did not have a single seizure for at least one year before T3. Of all patients operated, 32% had remained continuously seizure free since surgery, 9% gained late seizure control after having had seizures at the one year follow-up (T1), 24% had a seizure relapse after having been seizure free at T1, and in 35% of the patients seizures were never fully controlled.
Employment rates at baseline were 79% in the operated and 77% in the non-operated group. At T3 59% vs. 56% were employed.
All patients underwent baseline memory testing (T1) and a long term follow-up of > 5 years (T3), 137 of the 161 operated patients had a postoperative evaluation at one year after surgery (T2). All 369 patients had verbal memory data available at T1 and T3, figural memory data were available for 359 patients, and executive functions were available for 253 patients (141 non-operated, 112, operated). For the 161 operated patients with an assessment at T2 verbal memory, figural memory, and executive functions data were available.
3.1 Cognitive outcomes
3.1.1 Changes from baseline (T1) to long term (T3)
For verbal memory, the results revealed significant interaction effects of “surgery” (F = 10.02, p = 0.002), “seizure control” (F = 9.81, p = 0.002), and “age” (F = 8.45, p = 0.004). Accordingly, better outcome was seen when patients were not operated, seizure free, and when they were younger. It is important to note that the age effect did not differ between operated and non-operated patients, and that in particular the retest interval had neither a general nor any differential effect on verbal memory. (Fig. 1)
As regards figural memory, only the factor “age” (F = 8.011, p = .005) had an effect on performance change over time. Accordingly, better outcomes were seen with a younger age. Again, no differential effect of age on performance change in operated versus non-operated patients was obtained, nor was the retest interval relevant for performance change.
For executive functions, analyses showed significant effects of “drug change” (F = 5.642, p = 0.018), “age” (F = 3.97, p = 0.049). Accordingly, favourable outcomes were seen with a reduced drug load and with a younger age. “Seizure control” missed to become significant in the multivariate analysis (F = 3.5, p = 0.06). In separate analyses, however, patients in both treatment groups improved in executive functions when they were seizure free (see Fig. 2). The retest-interval had no effect on change in executive functions, neither alone, nor in interaction with other variables.
3.1.2 Changes from standard one year follow-up (T2) to long term (T3)
The same MANOVAs were calculated again, comparing the cognitive course of operated patients between T2 and T3 with the course of the non-operated controls between T1 and T3.
For verbal memory change, significant effects of “age” (F = 7.0, p = 0.008) “seizure control” (F = 4.9, p = 0.028″), “surgery” (F = 4.7, p = 0.031) and “drug change” (F = 4.4, p = 0.038) were observed, i.e. improvements were associated with a younger age, controlled seizures, no surgery, and reduction of drug load.
For figural memory change only the effect of “age” (F = 6.3, p = 0.012) reached significance. Thus improvements were seen in younger patients.
For executive functions the results indicated a significant effect only of the repeated factor “retest” (F = 4.6, p = 0.03). The effects of “drug change”, “age”, “seizure control”, and “surgery” did not reach the significance at the 95% level (all p < 0.1).
3.1.3 Individual level analyses
Individual level analyses indicate baseline impairments (T1) in executive functions or memory (verbal or figural) in 46–54% of the non-operated patients, in 51–61% of the surgically treated left temporal and 32–59% of the right temporal patients.
As regards performance change in executive functions, verbal memory, or figural memory from baseline (T1) to the long-term follow-up (T3), the non-operated group improved in 14–26% of the cases when patients were seizure free as compared to 10–12% when seizures persisted. Decline was evident in 0–6% seizure free patients as compared to 10–20% of those who were not seizure free (Table 2 a/b/c/).
Table 2Average performances, individual impairments, and individual performance changes over time.
a) Non-operated control group
b) Left temporal lobe surgery
c) Right temporal lobe surgery
Δ T1 T2
21% ↓ 10% ↑
34% ↓ 5% ↑
4% ↓ 4% ↑
19%↓ 8% ↑
Δ T2 T3
3% ↓ 17% ↑
11% ↓ 5% ↑
7% ↓ 7% ↑
11%↓ 11% ↑
Δ T1 T3
3% ↓ 20% ↑
12% ↓ 12% ↑
17% ↓ 20% ↑
37% ↓ 4% ↑
10% ↓ 16% ↑
20%↓ 9% ↑
Δ T1 T2
17% ↓ 14% ↑
17% ↓ 10% ↑
7% ↓ 3% ↑
22%↓ 14% ↑
Δ T2 T3
0% ↓ 21% ↑
10% ↓ 15% ↑
7% ↓ 7% ↑
14%↓ 18% ↑
Δ T1 T3
6% ↓ 14% ↑
20% ↓ 10% ↑
17% ↓ 17% ↑
20% ↓ 12% ↑
7% ↓ 13% ↑
17%↓ 13% ↑
Δ T1 T2
5% ↓ 14% ↑
0% ↓ 6% ↑
0% ↓ 20% ↑
0%↓ 1% ↑
Δ T2 T3
0% ↓ 5% ↑
14% ↓ 3% ↑
5% ↓ 5% ↑
12%↓ 4% ↑
Δ T1 T3
0% ↓ 26% ↑
7% ↓ 10% ↑
0% ↓ 24% ↑
12% ↓ 6% ↑
5% ↓ 19% ↑
8%↓ 11% ↑
M/SD = mean/standard deviation.
T1 T2 T3 = baseline, one year postoperative follow up, long term follow up.
For the individual level follow-up analyses of surgical patients, the patients were divided into a left and right surgery group. Assessments at baseline (T1), one year postoperatively (T2) and long term (T3) were taken into consideration.
The results indicate that in the operated group major decline was already evident at the one year postoperative follow-up evaluation. When patients were not seizure free at T3, 34% left and 19% right temporal resected patients significantly declined in verbal memory as compared to only 21% left and 4% right resected with decline in the seizure free group. Improvements in verbal memory were rare with 4–10% surgically treated patients (left and right resected) showing improvement at the one year follow-up. In figural memory, 7–22% of patients operated on the left or right showed a loss at the one-year follow-up. Gains in figural memory were evident in 3–14%.
In the time interval between the one year follow-up (T2) and long-term follow-up (T3) 17% of those who had undergone left temporal resections and 7% of those operated on the right improved in their verbal memory performance. Decline in the left vs. right surgery groups were seen in 3% and 7% of the seizure free patients versus 11% of those not seizure free. In terms of figural memory, 21% of patients categorised as seizure free after left versus 7% after right temporal lobe surgery improved, but improvements were also present when seizures were not controlled (15% vs. 18%). None of the left vs. 7% of the right temporal seizure free patients declined, as compared to 10% and 14% of the left or right temporal patients who continued to have seizures.
The question whether surgery accelerates memory decline can be answered by determining the number of patients who declined in the time interval between the first follow-up one year after surgery (T2) and the long term follow up assessment (T3). Overall about 9% demonstrated significant losses in verbal memory, figural memory, or executive functions over the T2-T3 interval. In the non-operated group 10%, 17%, and 6% showed a decline in verbal memory, figural memory, or executive functions between T1 and T3. The difference in performance changes between the two groups was not significant (Chi2 = 1.4, 1.8 and 4.3 df4, p = 0.8, p = 0.7, p = 0.4). The result did not change when subgroups created on the basis of seizure control were examined separately.
Comparing the individual long term changes in operated (T1-T3) and non-operated patients (T1-T3), the only significant difference was seen for losses in verbal memory in those operated on the left and failing to achieve seizure freedom (Chi2 = 18.0, df 4, p = 0.001)
In the analyses reported above, losses may be underrated if patients with pre-existing impairments who could hardly decline (“floor effect”) are included in the analysis. Therefore, the same analysis was repeated after exclusion of those who were already impaired at baseline T1, or at T2 respectively. After correction, the overall rate of patients with losses maximally increased by 12%. In the high-risk group of those operated on the left and not achieving seizure control the number of patients who lost between T1 and T2 increased from 34% to 46%, (seizure free: 21% to 31%), and the proportion of those who lost in the interval T2 to T3 increased from 10% to 15% (seizure free; 0% to 11%).
Another question of importance was to look at the number of repeated losers. It is remarkable that within the group of patients treated surgically only one patient lost repeatedly (i.e. between T1 and T2 as well as between T2 and T3) in verbal and another in figural memory. Similarly, only two “double winners” were identified in the respective domains. The majority of patients either lost or gained in memory in the year after surgery or during subsequent follow-up.
3.1.4 Age and ageing
A major aspect of this study was the question of whether age and the retest interval negatively affect the course of cognition after surgery. MANCOVAs in general revealed nonspecific age effects (worse performance with increasing age, or better performance with a younger age). There was no effect of passing time, i.e. of the retest-interval on outcome in memory or executive functions.
Dividing the groups by median split into two groups (younger vs. older than 30 years) and calculating correlations of performance changes (losses) to the retest interval generally revealed negative correlations in the younger and the older group as well (i.e. the longer the interval the better). The correlations were significant for executive functions (r= -0.234, p = 0.009) in the younger group. In older patients none of the correlations reached statistical significance.
3.1.5 Vocational outcome
Vocational outcome was assessed at T3 and related to the condition at T1. No significant difference between operated and non-operated patients was revealed at any time in terms of the number of patients in employment. Accordingly, there was no significantly different change over time (still employed: 54% non-operated, 65% operated, still in education: 21% non-operated, 17% operated, gain in employment: 3% non-operated, 4% operated, loss of employment: 22% non-operated, 23% operated). Overall the tendency in terms of change of vocational status was negative (22%) rather than positive (3%).
Multiple stepwise regression with pairwise exclusion of cases indicated that the employment situation at baseline (t = 12.0, p < 001), executive functions (t = 3.1, p < 0.01) and chronological age (t=-2.0 p < 0.05) predicted changes in vocational status between T1 and T3, explaining 45% of the variance. Accordingly employment at baseline, good baseline executive functions, and younger age were positive predictors. Surgery, gender, seizure outcome, drug change, verbal or figural memory and the follow-up interval did not enter the prediction model.
Temporal lobe epilepsy surgery, although very successful in terms of seizure control, often causes additional memory decline within the first year after surgery. The question whether temporal lobe resected patients will experience further memory decline with long term follow-up intervals is still up for debate. Continuing seizures, normal, or even pathological ageing may contribute to this.
The main aim of the present study was the evaluation of the longer term outcome in memory performance after temporal lobe surgery with retest intervals between baseline and long term follow-up of between 5 and 22 years with a median interval of 7 years. The focus of the analysis lay on the impact of surgery (additional damage of memory structures), epilepsy (seizure control), drug change (drug load/number of AED) and of age (chronological age) and ageing (retest-interval) on the course of memory.
To control for the effect of surgery a heterogeneous group of non-operated, pharmacologically treated epilepsy patients with focal epilepsies served as a control group. Since temporal lobe surgery specifically affects memory functions, extratemporal non-memory executive functions served as a control condition on the functional level.
We need to emphasize that this is a retrospective and observational single centre study (Bonn/Germany), and that the generalizability of its findings is therefore limited. The patient selection is biased but presumably in a negative rather than positive direction. It is very likely that the more strongly impaired patients returned to a highly specialized epilepsy clinic for long-term follow-up (regardless of whether they had undergone epilepsy surgery or not), whereas others stayed with their local physician.
Seizure control served as an experimental condition rather than an outcome parameter. Samuel Wiebe´s randomized trial published in 2001 [
] indicated that 58% of patients in the group treated with temporal lobe surgery were seizure free one year after surgery versus 8% with medical treatment. Our own controlled long term follow-up study from 2003 showed that 63% surgically treated versus 12% purely medically-treated patients with temporal lobe epilepsy became seizure free after 2–10 years [
]. Although not controlled and biased, the present data again confirm what is generally accepted, that, in the long run, surgery controls seizures better (41% seizure free) than pharmacological treatment alone (17% seizure free).
The data also confirm the known risk of postoperative memory losses, especially verbal memory loss after left temporal lobe surgery. Elisabeth Sherman´s metastudy and our own monocentric study with large groups of patients who had undergone temporal lobe resections indicate that one year after left temporal surgery the losses in verbal memory are twice as high than after right temporal lobe surgery (∼ 40% vs. 20%) [
]. In the present sample, 34% left temporal and 19% right temporal patients showed losses in verbal memory one year after surgery when seizures were not controlled as compared to 21% and 4% when seizures were controlled.
Leaving the effect of surgery aside, the following finding of this study is of major importance: The individual changes from the one year postoperative assessment to the long-term follow-up in operated patients compared to the changes in the non-operated control group between baseline and long term follow-up revealed no disadvantage for the operated group. 5–22 years after surgery, and compared to baseline, only 17% of those who had undergone left and 10% of those who had right temporal lobe surgery showed losses in verbal memory when they were seizure free, as compared to 37% of patients after left and 20% after right temporal lobe surgery if their seizures continued.
Most remarkably, only two patients showed further losses when they had already experienced a memory loss at the one-year follow-up. This finding and the lacking difference between controls and operated patients (when only the time from one year postoperative to long-term follow-up is considered) are strong arguments against the position that surgery accelerates memory decline in the long run.
This quite positive, and not - as possibly expected - increasingly negative long term outcome in memory after surgery was confirmed by group analyses. As displayed in the figures, the group who had undergone left temporal resections and did not become seizure free is the one that significantly declines in verbal memory after surgery and fails to recover from this impairment in the long run. The factors which determine the long-term outcome from before surgery to the final follow-up and from the one year postoperative assessment to the final follow-up are largely the same.
Operated and non-operated patients show recovery from baseline or postoperative impairments in the long run if they are younger, become seizure free, and when the drug load is reduced. The results thus confirm the positive impact of seizure control on postoperative performance and extend that was indicated by our own and Baxendale’s long term studies with shorter retest-intervals previously [
An important result of this work is that age but not ageing (i.e. the retest interval) was negatively related to cognitive impairment and decline. The age effect, i.e. older patients perform worse, was to be expected since the test scores were standardized with regard to age. Age was instead explicitly chosen as a covariate. The age effect was evident in both treatment groups and regardless of the cognitive domain. In particular, there was no interaction effect of the retest-interval with surgery and there was not one significant result in regard to possible effects of the-retest interval on cognitive change. Contrary to what might have been expected, but consistent with the finding of recovery in the long run, longer intervals were related to greater improvement rather than greater deterioration. This was particularly true for executive functions and to a lesser degree for figural memory if patients were younger (age at surgery < 30 yrs.). This was not observed in the older group of which at long term follow-up (T3). The oldest patient operated was 68 and the oldest non-operated 77 years old. Thus, within this age range, age and ageing restrict the capacities for functional recovery rather than promoting progressive decline.
Keeping the age range in mind, the result is consistent with our study from 2003 and with the findings of the Swedish, Dutch, and British long term follow-up studies according to which no accelerated cognitive decline or premature ageing should be expected after temporal lobe epilepsy surgery. [
The major focus of this study was on operated patients, but it should not be forgotten that seizure control and drug load reduction also had positive effects on cognition in the non-operated group. Since the non-operated group was considered as control and no clinical differentiations were made we refrain from any further interpretations.
The quite positive outcome in cognition was unfortunately not reflected by the vocational outcome. It seems that even in seizure free patients with preserved or recovered cognition, gains in employment status can hardly be expected. Independent of surgery and seizure outcome, the job situation at baseline, executive functions and chronological age predicted the vocational status. In a German publication of our long-term outcomes in 2000 [
] very different results were obtained. In this study 40% of non-operated patients changed from employment into early retirement as compared to only 10% in the operated group. At that time seizure control, younger age, male gender, and preserved memory predicted a positive development in regard to work. Presumably the different findings express changes of the economic situation and job market over time in addition to insurance benefits which may support early retirement.
In summary, the data do not confirm concerns that patients undergoing temporal lobe epilepsy surgery are likely to develop accelerated memory decline over the longer term. In order to optimize cognitive outcomes, seizure freedom and reduction of drug load appear most essential. Both factors are partially in the hand of the treating physician. Choosing drugs with favourable cognitive side effect profile may help [
]. Even if it takes time, seizure free patients have the potential to recover from postoperative impairments. As has been discussed elsewhere, patients may well take the risk of additional impairment if they can expect seizure freedom [
] and it limits recovery. For children who have undergone temporal lobe resections it has been shown that, due to greater functional plasticity, recovery from postoperative impairment is possible even within the first year after surgery [
] can have a negative long term effect on cognition. These issues become increasingly relevant when operated patients reach an age when dementia becomes more likely and when surgery is increasingly performed in older patients with late onset temporal lobe epilepsies.
As pointed out before, the present study is observational and retrospective. However it is controlled in a sense that experimental contrasts were set post hoc. Because of this and because of the patients’ selection criteria the representativeness of the results is restricted. The patients are especially not representative for those who became seizure free and never return to the specialist epilepsy centre after their epilepsy surgery. In favour of this study this might be considered a negative selection bias rather than a positive one. In this regard the finding of mostly stable, if not improved, cognitive functions is especially encouraging.
Taking the shortcomings into consideration, the conditions of surgery versus no surgery, seizure free versus not seizure free, drugs reduced versus not reduced, younger versus older and left versus right temporal surgery, memory versus non-memory functions, short and long-term retest-interval, were well suited to answer the major question of this study: “Does memory continue to decline in the long run after unilateral removal of structures involved in memory, when taking into consideration the usual determinants of postoperative cognitive outcomes?” Patients who have undergone temporal lobe resections at an advanced age or very late onset epilepsies are missing, so nothing can be said about these groups. The study also did not take different types of surgery into consideration. In this regard it should be considered that prospective randomized trials with very long term follow-up are very unlikely in the future, that patients, aetiologies, and surgical treatments change over time, and that we have to deal with cohort effects, which means that results which apply for patients operated in a given period of time may not apply to patients operated in another period of time [
]. So studies spanning a long period of time can only identify gross effects and principles. Finally, calculating individual changes bears the risk of underestimating losses because of the so-called floor effect, i.e. those who perform poorly cannot lose [
]. Breaking patient groups further down into impaired and not impaired patients would have led to groups too small for meaningful statistical analyses. In this regard it should be mentioned that for the whole sample the rate of losers was somewhat higher when patients who had impairments at baseline were excluded (T1-T3 10% for verbal memory, 16% for figural memory, and 2% for executive functions).
Dr. Helmstaedter reports personal fees UCB Pharma, EISAI, and GW-Pharma, Assurance companies, Courts, Occupational insurance associations, grants from EU (E-Pilepsy/EAHC and EpiCARE), outside the submitted work.
Dr. Elger reports personal fees from Bial, personal fees from Cyberonics, personal fees from Desitin, personal fees from Eisai, personal fees from Novartis, personal fees from Pfizer, personal fees from UCB, outside the submitted work.
Ms. Vogt has nothing to disclose.
A randomized, controlled trial of surgery for temporal-lobe epilepsy.